On the potential contributions of hybrid intelligent approaches to Multicomponent Robotic System development

Information Sciences 180 (2010) 2635–2648

Contents lists available at ScienceDirect

Information Sciences journal homepage: www.elsevier.com/locate/ins

On the potential contributions of hybrid intelligent approaches to Multicomponent Robotic System development R.J. Duro a,*, M. Graña b, J. de Lope c a

Grupo Integrado de Ingeniería, Universidade da Coruña, Spain Grupo de Inteligencia Computacional, Universidad del País Vasco, Spain c Grupo de Percepción Computacional y Robótica, Universidad Politécnica de Madrid, Spain b

a r t i c l e

i n f o

Article history: Received 17 April 2009 Received in revised form 8 January 2010 Accepted 1 February 2010

Keywords: Intelligent robotics Multi-robot systems Modular robotics

a b s t r a c t The area of cognitive or intelligent robotics is moving from the single monolithic robot control and behavior problem to that of controlling robots with multiple components or multiple robots operating together, and even collaborating, in dynamic and unstructured environments. This paper introduces the topic and provides a general overview of the current state of the field of Multicomponent Robotic Systems focusing on providing some insights into where Hybrid Intelligent Systems could provide key contributions to its advancement. Thus, the aim is to identify prospective research areas and to try to delimit the field from the point of view of the following essential problem: how to coordinate multiple robotic elements in order to perform useful tasks. Ó 2010 Elsevier Inc. All rights reserved.

1. Introduction The classical concept of a general-purpose industrial robot, which can be either manipulators or mobile robots, makes sense when the task to be carried out takes place in static, controlled and completely structured settings. However, when the environments are highly dynamic and unstructured and when the tasks to be performed are seldom carried out in exactly the same way, additional properties are required from the robotic systems to be used. Examples of such environments are shipyards, plants for constructing unique or very large structures or civil engineering sites. Unlike traditional automated plants, in these settings the work is not carried out in structured production lines where the product is moved along a series of stations that perform different operations on it, but rather, a series of individuals or teams of specialized workers are moved to and over the product to perform the tasks. As a consequence of the increased mobility, dynamicity and uncertainty that arises from this type of operation and environments, innovative approaches based on stronger design specifications such as modularity, scalability, fault tolerance, ease of reconfiguration, low fabrication and maintenance costs and adaptation capabilities are required to permit the automation of the different processes involved. From an industrial optimization point of view, as this type of operation involves a multitude of different tasks and processes that are generally sequential and usually over a single structure, it makes sense to have a small group of robotic structures that can easily adapt their hardware and capabilities to each task as it comes up rather than many specialists that will move to the structure as they are required and spend most of the time waiting for their turn. At the same time, to make these robots capable of operating in dynamic settings, they must be endowed with functionalities that allow them to adapt morphologically, functionally, and operationally to their environment in real time. Finally, and from an operational efficiency * Corresponding author. Address: Grupo Integrado de Ingeniería, Universidade da Coruña, EPS, Mendizábal s/n, 15403 Ferrol (A Coruña), Spain. Tel.: +34 981337400x3281; fax: +34 981337410. E-mail addresses: [email protected] (R.J. Duro), [email protected] (M. Graña), [email protected] (J. de Lope). 0020-0255/$ - see front matter Ó 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.ins.2010.02.005

2636

R.J. Duro et al. / Information Sciences 180 (2010) 2635–2648

point of view, it would be desirable for them to continue to operate even after failures occur in some of their components, that is, they should degrade in a non-catastrophic way. All of these requirements lead to the need for Multicomponent Robotic Systems (MCRS) in terms of having several robotic components or modules that can be configured in multiple ways to address different tasks. To make MCRS feasible, the design of a modular architecture to provide a convenient standardization of the interfaces between modules and an appropriate organization of perception, processing and control for these types of systems becomes a necessity so that they can reconfigure and adapt in an intelligent manner in order to complete the mission with the highest possible level of autonomy. The notion of Multicomponent Robotic System (MCRS), which generalizes and refers to the diverse modular and multirobot systems we are interested in, is used throughout the paper. An atomic component of these systems will be referred to as an individual or module. It may be a functionally complete robot unit or a module that needs to be combined with others in order to produce a desired functionality. A MCRS is defined as a set of individuals with a superimposed architecture. This architecture consists of the definition of a spatial distribution for the individuals, of a set of local and global control algorithms, and of a collection of communications channels and protocols for the exchange of information among individuals. The terms global conditions or global properties refer to the whole set of individuals, whereas local conditions and properties refer to an isolated individual. In this paper MCRS will be considered from several key points of view, relating to the state of the art and trying to identify relevant Hybrid Intelligent Systems based avenues for research. Hybrid Intelligent Systems (HIS) are characterized by the composition of different Computational Intelligence tools (Bayesian reasoning, neural networks, fuzzy systems, statistical classifiers, evolutionary algorithms, etc.) in a way that is adapted to the particular problem to be solved. They are aimed at achieving the highest degrees of flexibility and adaptation. Until now, most MCRS are characterized by the simplicity of their control systems, which are often handcrafted for the particular task to be demonstrated. We believe there is a wide open application field for Hybrid Intelligent Systems based approaches in the development of innovative MCRS. The paper is organized as follows: Section 2 is devoted to the review of some desired properties for MCRS. Section 3 will deal with the coupling among individuals. Section 4 considers the morphology of systems and individuals. Section 5 is dedicated to the implications of the environment and the task definition. Section 6 discusses control issues. Section 7 refers to perception aspects of MCRS and, finally, Section 8 summarizes the ideas presented in this paper. 2. Some desired properties of MCRS The basic expectation about MCRS is that they must be endowed with specific properties that allow them to perform tasks that no individual robot is able to carry out by itself or to perform them with increased efficiency and economy. Some examples of this class of systems and how they are applied with different degrees of success, as described in the literature, will be provided throughout the paper. To establish some basic common ground in this problem, this section reviews the most relevant properties that should be the distinctive mark of a new MCRS generation, presenting, at the same time, some of the systems that are currently being proposed. Four aspects will be considered in order to organize this section: autonomous reconfiguration, autonomy, advanced human–robot interaction and, finally, issues of self-perception and self-localization. 2.1. Autonomous reconfiguration It is the ability of a system to modify the functional/spatial configuration of its individual components without external intervention. This property is the most critical one in the present state of the technology. In current state of the art systems reconfiguration is not autonomous and external (human based) interventions take place at two levels: at the point in which the decision on the need to perform the reconfiguration is made and during the planning of the reconfiguration process itself. That is, most of the times the determination of the goal configuration and the computation of the sequence of intermediate steps needed to reach this goal configuration are carried out outside the MCRS (in the sense that computations are not performed by any of the individual robotic components), in many cases even the programming of the traction/actuators to carry out the reconfiguration is performed ad hoc. The decision on the need to perform the reconfiguration of the MCRS implies the evaluation of the current MCRS configuration relative to the assigned task and the state of the environment [96]. In most current state of the art systems this decision is not obtained automatically except for some very low level tasks, such as pursuit, which do not require a specification of this decision or the reconfiguration process [100]. Some autonomous reconfigurations in simulated systems have been reported using cellular automata [161], however the state of the art is very far from feasible physical realizations. The idea that robots should be able to self-configure is introduced in [89,109,110,170,187] for one type of modular robot, the M-TRAN [78,88,89]. In the current state of its development the transitions between configurations are achieved manually. However, a few reported studies may be found where configurations for particular tasks are obtained through genetic algorithms and other global random search methods [188]. Providing reconfiguration capabilities implies two different types of problems. On one hand there is the computational problem of estimating how to reach the target configuration from the current one through a feasible series of operations. On the other it is necessary for the hardware of the system to allow for the physical reconfiguration. In the case of distributed robotic systems, to reconfigure is basically to change relative positions. However, in the case of modular robots, the different modules must be able to perform physical coupling and

R.J. Duro et al. / Information Sciences 180 (2010) 2635–2648

2637

decoupling operations autonomously. Consequently, there is a need for specific hardware designs that allow for these actions and which involve the mechanical aspects of the coupling mechanism, its actuation systems and the design of the communications and powering contacts to allow for autonomous coupling and decoupling. The body of work on this topic in the literature is very small and usually related to very simple toy systems. For instance, the idea of self-configuration over Yim’s polybots is presented in [185]. Regarding coupling methods Liping et al. [93] propose one for planetary robots that is based on position sensors, and the work presented in [121] studies one which is based on magnetic couplings that may lead to new proposals of self-configurable robots. In terms of the computational problem, an area of active research is that of the application of intelligent (hybrid) systems for the autonomous on-line reconfiguration of this type of robots. It would really be necessary to reformulate the problem as a distributed optimization problem with partial information and to combine estimation methods (Bayesian or neuronal approaches) with robust optimization methods (evolutionary or graduated non convexity (GNC) techniques) in order to solve it. 2.2. Autonomy It is a desirable general property for any type of robotic system in all kinds of situations, but especially in unstructured dynamic environments. Autonomy implies that the system is able to adapt its configuration and its perception and control systems to the changing environmental conditions without human intervention. Autonomy implies life-long learning [19,51,55,118]. It also implies self-sufficiency, meaning the ability to preserve its own life without external intervention (power and maintenance). It is also important to consider the autonomy of the individuals within the system. In other words, it is convenient to examine how degrees of freedom and responsibilities are passed down to the individual level when the whole system increases its autonomy. 2.3. Advanced human–robot interaction (HRI) MCRS must coexist and interact with human operators in the same environment. They must comply with security standards, and, consequently, deal with communications to/from human operators. Regarding communications from humans, the focus is on the task specification problem. It would be desirable to have systems able to correctly interpret incomplete and imprecise specifications, as discussed above. Moreover, when it is embedded in unstructured dynamic environments, the robotic system will be required to guess the information needed to adapt the task specification to the current environment state. The HRI may imply multimodal signal processing (image, sound, gestures, etc.) [64,66], but at present the logical/ symbolic process of ensuring that the task will be correctly performed regardless of its poor specification is of paramount importance. Another interesting feature of new generation MCRS will be their ability to autonomously perform the decomposition of the task according to their configuration, comprising the possible reconfigurations as a result of this task evaluation and decomposition. When considering cooperating systems, the task decomposition should be additive: The completion of the task results from the addition of the completed results of the subtasks. In the case of coordination, the task decomposition is additive along the time axis: there is a timing requirement or some kind of completion order for the subtasks. On the other hand, a maximization process is assumed for competition in which the same task is assigned to many individuals, which will provide the best result by competing among themselves, or the competition process will produce as a side effect some emergent behavior that solves the task [160]. 2.4. Self-perception and self-localization Self-perception means that the system should be able to determine (diagnose) its configuration from sensing information about effector states. Self-localization implies determining the relative positions of its robot units or modules, and their current functionalities. Additionally, the system should be able to provide its spatial position as a whole. It may imply the construction of a map of the environment [15], or simply the position of a single individual (such as the hose head in systems that transport hoses and are linked to them) relative to its desired position. The capability of simultaneous localization and mapping (SLAM) would provide the maximum autonomy to the system [113,117,138]. All of these properties sought for operating MCRS are a direct or emergent result of the design and programming decisions that go into building a system of this type. Thus, what follows tries to provide an overview of the different proposals found in the literature in order to achieve these objectives concentrating on the opportunities for future research provided by Hybrid Intelligent Techniques. The elements that influence the creation and operation of any MCRS will be divided into five types: coupling, morphology, environment and tasks/activities, control and perception. 3. Coupling As this paper is aimed at dealing with equivalent problems across the diverse types of MCRS, some way to create a taxonomy of the different kinds of systems involved would be required. This is achieved through a parameter related to the degree of coupling among individuals, or components, which characterizes the strength of their coupling. Three types of coupling may be identified: physical, functional and informational coupling. This section reviews different works reported in the literature related to MCRS along these three coupling axes and their implications.

2638

R.J. Duro et al. / Information Sciences 180 (2010) 2635–2648

Some of the large groups of MCRS found in the literature may be characterized by considering a gradation along the physical coupling axis. When there is no physical coupling whatsoever, we have an uncoupled distributed system. Examples are robot soccer teams [54,75,83], teams of UGV or UAV [115,132,133], or uncoupled swarms [139,159,186,190]. Within this category a lot has been written about robot swarms [44,47,164,165], that is, relatively large groups of physically uncoupled robots that collaborate to achieve an objective, for example, rescue tasks [159,160] or material handling tasks in flexible fabrication cells [46]. When a rigid physical coupling is considered between components producing a new unit with new physical and functional properties, we have what are generally called modular systems. Examples of these systems are the polybot [60], M-TRAN [89], Proteo [16], Uni-drive [79] or even, in some cases, the s-bots, which can be coupled to perform some specific tasks [45,87,107,108,143,173], being really a hybrid between modular and distributed systems. Modular robots present structural degrees of freedom in order to adapt to particular tasks. One of the first practical implementations of modular robots was a sewer inspection robot [33] although the idea of reconfigurable modular robots starts with the designs by Yim [175,182–185] of a polypod robot that is capable of adapting its structure in order to produce different gaits for moving over different terrains. In [29] this philosophy is applied to the design of flexible fabrication cells. Finally, a third level of physical coupling can be defined for systems that are coupled through a passive non-rigid element. These systems are called Linked System throughout the paper. The paradigmatic case is that of a hose manipulated by a team of robots attached to it, or grabbing it (that is, the hose becomes a part of the robot system). This connection imposes constraints on the robot dynamics, which interfere in their coordination. It is also a non-linear transmission medium for the dynamical influences among robots. It is important to note that MCRS are contemplated as a generic class of systems and the objective here is to discuss ways of providing them with global autonomous actuation and reconfiguration strategies that permit the creation of heterogeneous robotic systems with any proportion of the three extreme types defined above and which delimit the domain (modular, linked and uncoupled distributed systems). For instance, robots such as the s-bots can act as uncoupled swarms or as modular systems [173]. Another interesting field of research is that of continuum manipulators [76] which are a limit case of modular systems that become almost linked systems. For illustration purposes some examples of multicomponent systems that combine the three extreme types to different degrees are presented in Fig. 1. The functional coupling axis, on the other hand, deals with the level of functional dependence between individuals required for the MCRS to be able to accomplish an assigned task or function. This dependence manifests itself in several ways. It may be that the individuals need to be placed in precise relative spatial positions, with or without physical coupling (i.e., the composition of s-bots to overcome an obstacle), that they need to perform relative motions (i.e., UAV trajectories for a given mission) or that the individual functionalities are applied according to a plan [96]. This logical dependence can be rigid (when the individuals must comply with a condition in an exact manner) o relaxed (when a range of variability or uncertainty is allowed). The system will be fault tolerant when it can accomplish its task/function regardless of the failures of individuals to meet their objectives (schedules). In general, it is essential to take into account the global system task/function to consider the functional decomposition and its assignment to the individuals. The functional dependence decomposition problem can be posed as a dynamic programming problem, where the a priori knowledge about tolerances and uncertainties in the functional coupling must be dealt with by probabilistic or fuzzy reasoning methods [96]. Finally, there is a third axis of coupling in terms of information shared by the different components of the MCRS. This axis is closely linked to communications and is highly influenced by the communications technology, topology and protocol used to exchange information among individuals and is often called communications coupling. The topology is given by the definition of the communications links and their properties (bandwidth, noise, persistency, delays). In some publications the communications coupling is identified with the system architecture, because it conditions the feasibility of distributed control, and the communications links may imply the existence of physical couplings or some spatial conditions (wireless coverage). For instance, the existence of communications delays implies dynamical effects on the design of the control system at the local and global level. In general, and even more for distributed control systems, it is desired that the communications

Distributed

EXAMPLE 1: Group of 4 modular robots that are linked by pairs to two hoses and that must coordinate to make the hoses discharge their products simultaneously in a given location.

EXAMPLE 2: Modular robot with two of its modules linked to a bar along which it must move.

Linked

Modular Fig. 1. Instances of MCRS as mixtures of the elementary system types.

R.J. Duro et al. / Information Sciences 180 (2010) 2635–2648

2639

topology be complete, that is, that each individual be connected to any other individual, but it is often impossible to achieve in an economic way and, consequently, strategies to deal with this incomplete and even changing topologies must be sought. It has been recognized that communications among the parts of the system or members of the team [158] is one of the critical aspects of MCRS, especially in the case of uncoupled distributed systems. Communications among individuals are usually implemented using radio-links whose robust performance as communications channels among the working robots accomplishing tasks may be problematic [48]. This is especially true when operating in real industrial environments performing real tasks. Take into account that industrial environments usually involve lots of machinery that introduce all kinds of noise in the communications channels. In addition, environments such as those relating to ship construction sites are metallic (ships consisting of steel plates) and consequently act as electromagnetic shields, hindering communications. To overcome this issue, optical communications and systems based on optical/photoelectric sensors and lasers can be studied as an alternative [43]. In some specific application cases, the communications task must become one of the ‘‘survival” tasks of the system and, as such, the system must introduce it within its behavior repertoire assigning resources for its implementation (units that act as messengers, the construction of opportunistic ad hoc networks, etc.). In fact, one of the problems in self-reconfiguring systems is how to correctly set up and maintain communications through morphological change or spatial redistribution processes. Some work has been carried out in this line such as that in [58] where, in the framework of modular robotics, the authors propose a hybrid communications system that can connect on-demand to form arbitrary network topologies. A special case in terms of communications that are embedded in the behaviors of the systems is the family of systems whose communications link is through marks on the environment (stigmergy). Some authors have made use of this type of approach using the pheromone metaphor [20,35,123–125,129,146,166], as in the case of ant colony algorithms. In this case the MCRS is structuring the environment to obtain some control effect. Thus, MCRS systems can go from uncoupled in terms of communications or information coupling, where the robots or components do their own thing without any regard for what the rest of the system is doing apart from what they are perceiving (this is the case of many reactive multirobot systems [128]), to weakly coupled systems such as those based on stigmergy to strongly coupled MCRSs with very clear protocols and communications approaches that allow all the components to have almost complete information on the rest of the components and even provide for shared planning facilities. Summarizing, MCRS may be classified along three coupling axes: physical, functional or informational. The latter has to do with the need for communications in the joint operation of individuals in order to conform a successful MCRS. In this line, a range of systems may be designed that go from completely uncoupled systems that do not need to communicate to perform their tasks to very closely coupled systems that need to continuously and extensively communicate to be able to operate, with different channels and approaches to performing the communications process, including specific behaviors for communications. In terms of functional coupling, tightly coupled systems require a very reliable and well defined operation of the individuals for the system as a whole to be able to operate. On the other hand, loosely coupled systems can tolerate errors or malfunctions in the operations of individuals and still perform the system task. Finally, in terms of physical coupling, a distinction may be established between distributed, linked and modular systems. 4. Morphology Robot shape determines functionality, that is, the kind of tasks the robot can perform. In other words, the embodiment determines the intelligence of the robot. This is also true for MCRS, the morphology of the systems conditions their possible applicability and intelligence.The tasks and functions that an individual may assume within the MCRS are conditioned by its own morphology and that of the whole MCRS. Two classes of MCRS systems in terms of the morphology of their components may be established, depending on whether all the components have the same morphology (homogeneous systems) [72] or not (heterogeneous systems) [168]. Heterogeneous systems present the advantage of lower cost individuals and less redundancy in the functionality repertoire of the system. Homogeneous systems present the advantage of independence from specific individual performance, as the individuals are interchangeable and may assume all the required functionalities, leading to improved fault tolerance. The morphology of the MCRS as a unit comes from the consideration of the system’s configuration [127] in addition to the morphology of each component. This is especially true for modular systems, where the physical coupling of modules produces a new unit with shape and properties that depend on the spatial configuration of the components. In uncoupled distributed systems, the morphology of the MCRS is related to the spatial distribution of the individuals, which conditions some aspects of the system, such as the communications network and, in some cases, the degree of accomplishment of the assigned task [92,102,127]. Most of the work in the morphology and configuration of uncoupled distributed MCRS systems focuses on the control system and its interaction with the different environments taking the morphology of the modules as preset, consequently, this will be considered later on in the corresponding section. In linked systems the spatial distribution includes the passive link element and its physical properties, which may strongly condition the whole system behavior. The morphological design of the individual components [79] determines the extent of the whole MCRS configuration space and its functionality. In the case of modular systems the emergent properties of a configuration usually allow performing tasks that the modules cannot perform on their own. One of the current challenges in this field is the optimal morphological design of the module in order to be able to produce some desired emergent properties through its integration in the MCRS [137]. Evolutionary algorithms, and other random search strategies, have been applied to this endeavour in the framework of evolutionary robotics [111,147].

2640

R.J. Duro et al. / Information Sciences 180 (2010) 2635–2648

The work in [175] discusses the limitations of metamorphic robots based on cubic modules. Different modular configurations are being proposed even nowadays, examples are [25,60,64,79,171]. New classes of robots are introduced in [21] where Campbell et al. present robots that are configured as power buses while performing the assigned task. In [23] Carrino et al. describe modules for the construction of feed deposition heads in the generation of composite materials. The results reported in [30,150,179,181] study kinematics calibration methods and ways for obtaining the inverse kinematics and the dynamics of modular and reconfigurable robots in order to solve the problems introduced by tolerances in the fabrication of the modules. On the other hand, [18] introduces a methodology for the dynamic modeling of multi-robot systems that facilitates the construction of simulators to be used in order to accelerate the development of intelligent control systems through virtual experiments. Regarding the automated design of modular robots, some work has been carried out in the application of evolutionary algorithms that seek the minimization of a criterion based on the variety of the modules employed for a given task that is kinematically characterized [180] or on the mass, ability and workspace [91]. In [192] Zhang et al. provide a representation of the robot and the environment that permits the application of case based reasoning techniques to the design of a modular robot. For the automation of the design of the modular robot configurations, including the case of self-reconfigurable robots, [82] proposes a representation of the potential connection topologies among the modules. In a formal approach to the problem, Saidani [144] discusses the use of graph theory and cellular automata as a base for the development of design and selfreconfiguration algorithms. Despite the activity along this line, research is mostly based on pre-designing systems and not really on their real-time redesign or adaptation to new components. This type of systems will not be really autonomous until these design and analysis tasks are carried out in an autonomous and distributed manner over the robot modules themselves allowing for real time reconfiguration of the modules when necessary. Again, robust estimation and modeling methods, that are still not in general use, as well as other HIS techniques that will constitute some type of distributed cognitive mechanism for the MCRS are required. 5. Environment and tasks/activities The environment, which is also called the external configuration space, is the operational field of the MCRS. It encompasses the physical space as well as the diverse operational constraints on its performance, from electromagnetic noise interference on wireless communications up to illumination variations or the spatial distribution of obstacles. In a structured environment all the elements are determined within a known variability range. The system designer and its human operator can trust this information about the system’s state. Moreover, the structured environment often helps in accomplishing the task (i.e., mobile robot navigation using floor printed paths/marks). The environment for the robotic system may determine much of its design. For instance, a robot designed for operating on vertical surfaces [141] shows many specific traits that are not shared with other robots. The motivation and justification for the creation of advanced robotic solutions lies in the assumption of new working environments with unfavourable conditions for robot operation, such as very cluttered environments [149]. The environment may even impose conditions on the self-assembly process [63]. In unstructured dynamic environments, the knowledge about their shifting conditions is uncertain and the usefulness of any robotic solution is restricted in time. In the paradigmatic case of shipyards or civil construction sites, the environment is not suited for conventional autonomous robots and, on top of that, it is continuously changing as the construction proceeds towards its completion. So we have a non-stationary trend superimposed on the stochastic variations due to normal operation in the environment. For static unstructured environments [94], an initial environment-scouting phase must be enough to build the map that would be used in the operation phase. The need for maps is unavoidable in many tasks where purely reactive behavior would not be appropriate to accomplish them. For dynamic unstructured environments, there is an inescapable need for regular monitoring, looking for changes, and the systematic updating of environment information. Some kind of life-long learning is required to deal with this kind on environments. When an objective or goal is imposed on the robotic system, a task to be performed is being assigned/defined, i.e., target tracking, garbage collection, search, transport, coverage of an space, soccer playing, hunting, escorting [7,12– 14,42,45,62,65,68,75,81,84,112,136,142,148]. However, a system may show activity that does not address the satisfaction of an externally set objective. In other words, not all the activities of a MCRS necessarily serve the accomplishment of an assigned task; the system may be operating even when no task has been assigned to it. For instance, individuals may have inevitable needs (battery charge, environment map maintenance) that force them to maintain some activity. These needs can be understood as endogenous tasks. In addition to the uncertainty about the state of the environment, the task specification itself may be imprecise or incomplete. Imprecise in the sense that the goal is not precisely stated,for instance, when a navigation destination is set relative to an undefined environment condition. The task can be incompletely specified in the sense that some situations are left unspecified. This situation may be the most common one in unstructured dynamic environments. Some robust specification interpretation and evaluation mechanisms will be needed to cope with imprecise and/or incomplete specifications. The problems raised to achieve such ability offer a rich and mostly unexplored avenue of application of HIS. 6. Control Control is the part of the system able to produce a decision/action sequence to reach a desired goal given some information about the environment and the system itself. As indicated above, the statement of the goal implies the definition of a

R.J. Duro et al. / Information Sciences 180 (2010) 2635–2648

2641

task and the control system must establish the strategy to accomplish this task. The definition of control implies the ideas of intelligence and optimization. Intelligence defined as the ability to plan and execute action sequences. The control process becomes fully or partially an optimization process when the stated goal can be formulated as the global extreme (maximum or minimum) of an objective function [111]. In unstructured and dynamic environments, control processes must be able to detect and also recognize variations taking place in the environment in order to adapt the decision and planning processes [71]. The objective function can be the preservation of a relative spatial configuration [120] or the time required to accomplish the task [68]. The uncertainty in the environmental knowledge introduces uncertainty in the evaluation of the objective function and on the effect of the decisions taken as a result of its optimization. There are already a few instances of heuristic proposals of control strategies [154,189]. However, there is a broad avenue for innovations coming from the HIS field. For instance, new optimization (i.e., evolutionary) algorithms dealing with imprecise, non-stationary or noisy objective functions may be useful in this setting. Two control levels may be defined for MCRS systems: the individual component level (component or actuator control) and at the global system level (mission control). The most substantial difference is that, often, global system goals are fixed by the task statement, in other words, the task is assigned to the global system and the component control systems goals must be crafted so that when all the components are operating together this global goal is achieved. Hierarchical or sequential task/goal decompositions may be found, so that the local optimization performed by each agent contributes to the global optimum seeking process [104]. This is often achieved in terms of different utility based strategies derived from the agent literature that permit defining private utility functions that will lead to the desired global utility under the appropriate environmental conditions. An example may be found in the principled evaluation function selection procedure for evolving coordinated multi-robot systems developed by Agogino and Tumer [3]. It is also possible to extract the individual’s goals from a competitive or cooperative decomposition of the global task goal [80,127]. Probably the most interesting approach for the decomposition of the global goal in industrial or real world applications is the one based on cooperative decomposition [80]. Cooperation may be used in order to achieve common goals, that is, to divide the tasks, which is sometimes based on task allocation algorithms or load balances, to avoid conflicts while the task is being accomplished, or to make a collective decision [10,38,51,127,132,135,136,168,169,173]. Sometimes the basis for the cooperation is to obtain the maximum reward for the system as a whole. Task decomposition strategies have been implemented by means of ant colony optimization (ACO) [28,39,194,196] or reinforcement learning paradigms [9,53,74,103,167,178,195]. On the other hand, the competitive approach is applied mainly when the roles of other environmental elements are taken into account as in robot soccer [9,54,75,83,167], or when the system is inspired on a general multi-agent approach [74,106]. It is important to note that global performance depends on individual performance, which is especially true for cooperative task/goal decompositions. Individuals must be trustworthy in the sense that the global system should be able to trust them to accomplish their corresponding goals in a timely fashion. Alternatively, the global system may be designed as fault tolerant, achieving the global goals regardless of the failure of individuals, at least of part of them. Note that competitive systems will be inherently fault tolerant. Another important distinction may be made between centralized [13,41,67] and decentralized (distributed) [24,36,108,116,126,127,145,146,153,156,164] control schemes. In the former, a single individual computes the task/goal decomposition, assigns subtasks/subgoals to individuals and maintains information about the global system progressing to the global goal. In the latter, the fulfilment of the global goal emerges spontaneously (synergistically) from the independent fulfilment of the individual tasks/goals. Evidently, the individual tasks must have been formulated in an appropriate way and the global problem must be decomposable (a common research issue in agent literature) [51,104]. This decomposition may be achieved by negotiation as in [130,145], by some form of swarm dynamics [190], or even by some form of learning [17]. The communications structure is intimately related to control decentralization, but does not necessarily determine it. For instance, it is possible to find decentralized control systems possessing a central communications node that performs the role of intermediate station, relaying messages between individuals. The control scheme must take into account the communications structure (topology, noise, reliability), although there are exceptional systems that accomplish cooperation without communication [140]. Decentralized control systems are expected to be more fault tolerant. The control schemes must constantly maintain updated information on the state of the environment, the system’s own state and the level of accomplishment of the task. There are two natural information levels in MCRS: individual and global. In centralized control schemes a single individual maintains the global information that supports/implements the central control. The fusion of local information takes place at the computational resources of this singular individual when required. In decentralized control schemes, the local information of each individual usually includes information generated by its own sensors and computational processes as well as an estimation of the local information in the remaining individuals, obtained from the communications received [190]. The study of consensus [132] heuristic algorithms and processes aims to establish the convergence of the local information under noise and uncertainty to a global precise picture of the whole system, environment and task state comparable to the one that a centralized controller can build with perfect information. Some authors emphasize the role of the communications network as the key for distributed control system development and deployment. We single out [132,133] who devote their efforts to the distributed control of autonomous vehicles performing a task cooperatively. Their central idea is the local estimation of a trustworthy local representation of the global system and environment states. The consensus building processes attempt to produce this representation in a robust way from the communications among individuals. The control of heterogeneous groups of robots may require the coexistence of heterogeneous control rules (set at the start) at the individual level [168] and the coordination between the different groups as a whole.

2642

R.J. Duro et al. / Information Sciences 180 (2010) 2635–2648

Some modular systems, such as the ones in [73,98,105], are better suited to be controlled in a feedforward manner by pattern generators that can be hybridized with neuronal structures. In fact, one of the M-TRAN self-reconfiguration schemes involves a central pattern generator and a genetic algorithm [89]. Due to the nature of this problem, it is very usual to find biologically inspired approaches for controlling the movements of this type of robotic mechanisms [95,151,157,191]. Usually they employ different kinds of hybrid approaches in order to generate locomotion patterns to be applied to the mechanisms. New HIS approaches for distributed control may be inspired by the hybridization of reinforcement learning with evolutionary algorithms [99], but also by most classical mixtures of declarative, procedural knowledge and case based reasoning [163]. The enrichment of situational calculus in [54] with other Computational Intelligence tools, may enable this approach to be extended to very unstructured dynamic environments. Distributed control can also be obtained by evolving fuzzy [174] or neuronal [75] controllers that exhibit some emergent behavior. The biological foundations of the idea of robot swarms are reviewed in [166], including a prospective of their application in [142]. In [57] Fukuda et al. discuss the advantages and disadvantages of multiagent robotic systems as compared to single robots. The individuals considered are in general very simple in their internal dynamics and, consequently, the introduction of sophisticated approximate reasoning systems would permit an extension of the range of behaviors and their robustness to changing situations. In [126] Peleg presents a universal architecture for the decentralized control of groups of robots. A review of the state of the art of decentralized control is given in [153]. Wessnitzer and Melhuish [177] integrate behavior based control strategies with swarm control systems in a task having to do with the elimination of underwater mines. Dorigo within his swarmbots project [45] presents a hunting behavior as a collective decision making process. In general, the formulation of decentralized control implies the need to work with incomplete or temporally inconsistent information. HIS should help to improve the robustness of these control systems. Finally, many design tasks, covering control, morphology and perception aspects, fall within the field of Evolutionary Robotics due to the fact that they can be formulated as the optimization of a given objective function [44,90,109]. As such, they are open to further improvement through HIS type approaches. 7. Perception Perception is inherent to any robotic system in order to obtain feedback information on the effects of the decisions taken and the actions performed. Perception subsystems include the sensors as well as the computational processes (filtering, feature extraction, encoding) that produce the information in the format used by the control schemes [49]. Sometimes, perception is the driving task [135] for the behavior of a MCRS. Robot perception can be seen from two perspectives, the perception of the robot’s own being (proprioception in biological systems) which is called self-perception here and some authors call self-recognition [59], and the perception of the environment. A part of the life-long learning processes required for systems embedded in unstructured dynamic environments consist in the adaptation of the perception subsystems. This adaptation ranges from the need to operate robustly under changing sensing conditions (i.e., changing lighting conditions for optical cameras) up to the need to discover new semantics (new objects or structures in the environment). Quantization and encoding are the ways to assign meaning to sensor data through the discovery and identification of information quanta equivalent to semantic concepts. Quantization partitions the data space to obtain a dimensional reduction or to remove uncertainty and noise. The association of discrete values to the partitions produces a data encoding. This data encoding allows translating the sub-symbolic data into the symbolic domain. Most of the adaptation processes in the perception subsystem are related to learning activities when training supervised and unsupervised quantizers/encoders, with some life-long learning versions for dynamic environments. Artificial Neural Network architectures [2,4,52,152] provide on-line clustering techniques to construct this life-long learning process. New proposals for quantization techniques lie in the frontier between the already established computational domains of artificial neural networks and fuzzy systems [61,134]. Some or all of the individuals within a MCRS may carry out perception tasks and, thus, a need arises for the integration of information coming from different locations at different times in order to provide a global usable view. This problem can be basically stated as a problem of merging the different sensory maps created by each component [1,5,15]. It can also be interpreted as a distributed AI problem where each agent carries out a part of the modeling task [56,155,169]. There are also approaches that include specific hybrid artificial intelligence methods for building the environment model [71,193].The fusion of the local perceptions (views) produces global information that can take the shape of environment maps [135] or other kinds of sensorial maps, which are sometimes modulated by attention processes. This fusion can be modelled as multi-agent systems [122] and may involve different sensors, that is, the sensing of the robot may need to be organized so that the different sensors are integrated in order to obtain the desired information. A primitive example is the application of Bayesian decision theory for door detection as presented in [86]. A decentralized Bayesian decision algorithm that may be used for the fusion of sensorial information in sensor networks is introduced in [97]. Map building is an essential perception process in many applications [1,5,31,56,70,135,169,172,176,193]. For instance, Kumar and Sahin [87] consider the problem of generating cognitive maps in mine detection and Pack and Mullings [116] introduce metrics for measuring the success of a joint search performed by a swarm and a universal search algorithm. A variety of techniques have been used to perform this task with distributed multi-robot systems, ranging from occupancy grids [1], to genetic algorithms [193], morphological neural networks [176], and particle filters [70]. The maximum expression of autonomy is the ability to autonomously perform simultaneous mapping and localization of the robot units (SLAM)

R.J. Duro et al. / Information Sciences 180 (2010) 2635–2648

2643

[11,38,40,41,50,172,176]. This is a promising avenue of research where hybrid techniques can be useful to obtain more efficient exploration and feature extraction algorithms. Kalman Filters have been applied with some success to this problem [6,8,22,26,37,67,69,77,114] in the context of SLAM. There are already some hybridizations of Kalman Filter based approaches with fuzzy systems for single mobile robots in environments that do not fit in the linear modeling paradigm easily [27,34,69,101,117,119,120,131]. Hybridization of the Kalman Filter approaches with Bayesian, fuzzy or neural network approaches may lead to interesting HIS avenues for research in the framework of MCRS. In MCRS, self-perception implies the estimation of the spatial configuration and state of the components. In modular systems, accounting for the module connections and the odometry of the attached actuators may be enough to build an accurate representation of the system configuration. For other kinds of systems, self-perception may imply processing sensorial information, like recognizing flags or symbols attached to individuals [85,100,132]. For uncoupled distributed systems the individual component position estimation may be carried out using a centralized (such as the zenithal cameras in robot soccer matches) or a distributed approach, where each individual recognizes other individuals belonging to the system and positions them. For linked systems, self-perception implies modeling and estimating the parameters of the passive linking element , in addition to the estimation of the individual robotic units’ positions. This process could be realized in a centralized or distributed manner through some consensus heuristic [132]. Finally, perception is intimately linked to control, making it very difficult in some instances to differentiate them within a system. One very interesting case is that of sensori-motor maps, usually implemented through artificial neural networks [98]. These maps provide fast trainable reactive responses, and can be mixed with other control loops. An interesting application is that of positioning and map generation through robot swarms [138]. A precedent may be found in [32]. In the same vein, but at a higher cognitive level, Stroupe and Balch [162] try to estimate the best next move of a group of robots in order to obtain the map. Summarizing, perception subsystems involve computational processes that require adaptability and life-long learning abilities for increased system autonomy. They offer a whole set of challenges for the development of new intelligent algorithms. Most of them will fall within the field of HIS. In fact, many current perception subsystems employ computational techniques based on variations of Kalmann Filters, which offer a broad spectrum of possible hybridization combinations. 8. Concluding remarks This paper has introduced some of the current issues within the field of Multicomponent Robotic Systems with the objective of showing where Hybrid Intelligent Systems could provide innovative contributions to the field. The basic focus has been placed on the fundamental problem of how to coordinate multiple robotic elements in order to perform useful tasks. After an initial statement describing the desired properties that every MCRS should posses, the main topics in the field have been addressed, including the coupling among the system components, the implications of the morphology of the whole system and also of its individuals, the influence of the environment and the definition of the tasks, and some issues on control and perception that are particular to this kind of systems. A broad spectrum of potential applications of HIS has become evident from this analysis. Algorithms are needed for the identification of the current state of the MCRS, and to decide its actions [70] in a collective distributed way, including the need to reconfigure the system. It is also necessary to develop efficient planning algorithms for the reconfiguration of a wide variety of systems, ranging from modular to uncoupled. New techniques for morphological design may allow fitting individual morphologies into the global functionality required to perform an assigned task. A systematic need of working with imprecise and incomplete information, which may be temporally inconsistent, is detected when contemplating distributed implementations of control and sensing. Sensor fusion, whether from several sensors from the same robot or from different robots, requires robust and efficient modeling techniques. As a final comment, we must say that the quest for MCRS that are useful for applications in real unstructured and dynamic environments is still in its infancy. Even though there has been a lot of work carried out in the last decade in this line, there is still a lot more left for these systems to achieve industrial level performance. This quest must necessarily involve an increase in the operational autonomy of the systems and we believe that one of the most promising paths to this objective is through the application of HIS based techniques. Acknowledgement This work was funded by the Spanish MEC through project DPI2006-15346-C03 (in part with ERD Funds). References [1] N. Adluru, L.J. Latecki, M. Sobel, R. Lakaemper, Merging maps of multiple robots, in: Proceedings of the 19th International Conference on Pattern Recognition (ICPR 2008), 2008, pp. 1–4. [2] B. Aisa, B. Mingus, R. O’Reilly, The emergent neural modeling system, Neural Networks 21 (8) (2008) 1146–1152. [3] A. Agogino, K. Tumer, Efficient evaluation functions for evolving coordination, Evolutionary Computation 16 (2) (2008) 257–288. [4] H. Amadou Boubacar, S. Lecoeuche, S. Maouche, SAKM: self-adaptive kernel machine a kernel-based algorithm for online clustering, Neural Networks 21 (9) (2008) 1287–1301. [5] F. Amigoni, S. Gasparini, M. Gini, Building segment-based maps without pose information, Proceedings of the IEEE 94 (7) (2006) 1340–1359.

2644

R.J. Duro et al. / Information Sciences 180 (2010) 2635–2648

[6] G.C. Anousaki, K.J. Kyriakopoulos, Simultaneous localization and map building of skid-steered robots, IEEE Robotics & Automation Magazine 14 (1) (2007) 79–89. [7] G. Antonelli, F. Arrichiello, S. Chiaverini, The entrapment/escorting mission, IEEE Robotics & Automation Magazine 15 (1) (2008) 22–29. [8] L. Armesto, J. Tornero, M. Vincze, On multi-rate fusion for non-linear sampled-data systems: application to a 6D tracking system, Robotics and Autonomous Systems 56 (8) (2008) 706–715. [9] M. Asada, E. Uchibe, K. Hosoda, Agents that learn from other competitive agents, in: Proceedings of the Machine Learning Conference Workshop on Agents That Learn from Other Agents, 1995, pp. 1–7. [10] J. Baber, J. Kolodko, T. Noel, M. Parent, L. Vlacic, Cooperative autonomous driving: intelligent vehicles sharing city roads, IEEE Robotics & Automation Magazine 12 (1) (2005) 44–49. [11] T. Bailey, H. Durrant-Whyte, Simultaneous localization and mapping (SLAM): Part II, IEEE Robotics & Automation Magazine 13 (3) (2006) 08–117. [12] B. Bethke, M. Valenti, J.P. How, UAV task assignment, IEEE Robotics & Automation Magazine 15 (1) (2008) 39–44. [13] R.M. Bhatt, C.P. Tang, V.N. Krovi, Formation optimization for a fleet of wheeled mobile robots – a geometric approach, Robotics and Autonomous Systems 57 (1) (2009) 102–120. [14] A. Bicchi, A. Danesi, G. Dini, S. La Porta, L. Pallottino, I.M. Savino, R. Schiavi, Heterogeneous wireless multirobot system, IEEE Robotics & Automation Magazine 15 (1) (2008) 62–70. [15] A. Birk, S. Carpin, Merging occupancy grid maps from multiple robots, Proceedings of the IEEE 94 (7) (2006) 1384–1397. [16] H. Bojinov, A. Casal, T. Hogg, Multiagent control of self-reconfigurable robots, Artificial Intelligence 142 (2002) 99–120. [17] A. Bonarini, V. Trianni, Learning fuzzy classifier systems for multi-agent coordination, Information Sciences 136 (1–4) (2001) 215–239. [18] C.S. Bonaventura, K.W. Jablokow, A modular approach to the dynamics of complex multirobot systems, Robotics, IEEE Transactions on Robotics 21 (2005) 26–37. [19] D. Brun-Picard, J.S.S. Sousa, Design of machines and robots endowed with a permanent learning ability, Control Engineering Practice 7 (1999) 565– 571. [20] P. Caamaño, J.A. Beccerra, R.J. Duro, F. Bellas, Incremental evolution of stigmergy-based multi-robot controllers through utility functions, LNCS, vol. 4693, Springer-Verlag, 2007, pp. 1187–1195. [21] J. Campbell, P. Pillai, S.C. Goldstein, The robot is the tether: active, adaptive power routing for modular robots with unary inter-robot connectors, in: Proceedings of the Intelligent Robots and Systems (IROS 2005), 2005, pp. 2960–2967. [22] M. Cao, L. Yu, P. Cui, Simultaneous localization and map building using constrained state estimate algorithm, in: Proceedings of the 27th Chinese Control Conference (CCC 2008), 2008, pp. 315–319. [23] L. Carrino, W. Polini, L. Sorrentino, Modular structure of a new feed-deposition head for a robotized filament winding cell, Composites Science and Technology 63 (15) (2003) 2255–2263. [24] R. Cassinis, G. Bianco, A. Cavagnini, P. Ransenigo, Strategies for navigation of robot swarms to be used in landmines detection, in: Proceedings of the 1999 Third European Workshop on Advanced Mobile Robots (Eurobot’99), 1999, pp. 211–218. [25] A. Castano, A. Behar, P.M. Will, The Conro modules for reconfigurable robots, IEEE/ASME Transactions on Mechatronics 7 (4) (2002) 403–409. [26] J.A. Castellanos, J. Neira, J.D. Tardos, Multisensor fusion for simultaneous localization and map building, IEEE Transactions on Robotics and Automation 17 (6) (2001) 908–914. [27] A. Chatterjee, F. Matsuno, A neuro-fuzzy assisted extended kalman filter-based approach for simultaneous localization and mapping (SLAM) Problems, IEEE Transactions on Fuzzy Systems 15 (5) (2007) 984–997. [28] A. Chatty, I. Kallel, A.M. Alimi, Counter-ant algorithm for evolving multirobot collaboration, in: Proceedings of the 5th international Conference on Soft Computing As Transdisciplinary Science and Technology, 2008, pp. 84–89. [29] I.M. Chen, Rapid response manufacturing through a rapidly reconfigurable robotic workcell, Robotics and Computer-Integrated Manufacturing 17 (3) (2001) 199–213. [30] I.M. Chen, Y. Guilin, Inverse kinematics for modular reconfigurable robots, in: Proceedings of the 1998 IEEE International Conference on Robotics and Automation, vol. 2, 1998, pp. 1647–1652. [31] C.M. Clark, Probabilistic road map sampling strategies for multi-robot motion planning, Robotics and Autonomous Systems 53 (2005) 244–264. [32] W.W. Cohen, Adaptive mapping and navigation by teams of simple robots, Robotics and Autonomous Systems 18 (1996) 411–434. [33] S. Cordes, K. Berns, M. Eberl, W. Ilg, R. Suna, Autonomous sewer inspection with a wheeled, multiarticulated robot, Robotics and Autonomous Systems 21 (1997) 123–135. [34] J. Dah-Jing, W. Sheng-Hung, Adaptive fuzzy strong tracking extended kalman filtering for GPS navigation, IEEE Sensors Journal 7 (5) (2007) 778–789. [35] T.S. Dahl, M. Mataric, G.S. Sukhatme, Multi-robot task allocation through vacancy chain scheduling, Robotics and Autonomous Systems 57 (6–7) (2009) 674–687. [36] I. Dahm, M. Hebbel, M. Hulsbusch, J. Kerdels, W. Nistico, C. Schumann, M. Wachter, Decentral control of a robot-swarm, in: Proceedings of the Autonomous Decentralized Systems (ISADS 2005), 2005, pp. 347–351. [37] S.J. Davey, Simultaneous localization and map building using the probabilistic multi-hypothesis tracker, IEEE Transactions on Robotics 23 (2) (2007) 271–280. [38] M. Di Marco, A. Garulli, A. Giannitrapani, A. Vicino, Simultaneous localization and map building for a team of cooperating robots: a set membership approach, IEEE Transactions on Robotics and Automation 19 (2003) 238–249. [39] Y. Ding, M. Zhu, Y. He, J. Jiang, An autonomous task allocation method of the multi-robot system, in: International Conference on Control, Automation, Robotics and Vision, 2006, pp. 1–6. [40] M.W.M.G. Dissanayake, P. Newman, S. Clark, H.F. Durrant-Whyte, M. Csorba, A solution to the simultaneous localization and map building (SLAM) problem, IEEE Transactions on Robotics and Automation 17 (3) (2001) 229–241. [41] K.D. Do, J. Pan, Nonlinear formation control of unicycle-type mobile robots, Robotics and Autonomous Systems 55 (3) (2007) 191–204. [42] R.L. Dollarhide, A. Agah, Simulation and control of distributed robot search teams, Computers & Electrical Engineering 29 (5) (2003) 625–642. [43] M. Dongtang, W. Jibo, Z. Zhaowen, A novel optical signal detecting and processing method for swarm robot vision system, in: Proceedings of the IEEE International Conference on Robotics, Intelligent Systems and Signal Processing, vol. 2, 2003, pp. 1081–1085. [44] M. Dorigo, V. Trianni, E. Sahin, R. Gross, T.H. Labella, G. Baldassarre, S. Nolfi, J.-L. Deneubourg, F. Mondada, D. Floreano, L.M. Gambardella, Evolving self-organizing behaviors for a swarm-bot, Autonomous Robots 17 (2-3) (2004) 223–245. [45] M. Dorigo, E. Tuci, R. Gross, V. Trianni, T.H. Labella, S. Nouyan, C. Ampatzis, J.-L. Deneubourg, G. Baldassarre, S. Nolfi, F. Mondada, D. Floreano, L.M. Gambardella, The SWARM-BOTS project, LNCS 3342 (2005) 31–44. [46] K.L. Doty, R.E. Van Aken, Swarm robot materials handling paradigm for a manufacturing workcell, in: Proceedings of the IEEE International Conference on Robotics and Automation, vol. 1, 1993, pp. 778–782. [47] G. Dudek, M. Jenkin, E. Milios, D. Wilkes, A taxonomy for swarm robots, in: Proceedings of the Intelligent Robots and Systems’93 (IROS’93), vol. 1, 1993, pp. 441–447. [48] T.W. Dunbar, J.M. Esposito, Artificial potential field controllers for robust communications in a network of swarm robots, in: Proceedings of the ThirtySeventh Southeastern Symposium on System Theory (SSST’05), 2005, pp. 401–405. [49] R.J. Duro, J.A. Becerra, J. Santos, Behavior reuse and virtual sensors in the evolution of complex behavior architectures, Theory in Biosciences 120 (2001) 188–206. [50] H. Durrant-Whyte, T. Bailey, Simultaneous localization and mapping: Part I, IEEE Robotics & Automation Magazine 13 (2) (2006) 99–110. [51] M.G. Earl, R. D’Andrea, A decomposition approach to multi-vehicle cooperative control, Robotics and Autonomous Systems 55 (2007) 276–291. [52] M.J. Er, Y. Zhou, Automatic generation of fuzzy inference systems via unsupervised learning, Neural Networks 21 (10) (2008) 1556–1566.

R.J. Duro et al. / Information Sciences 180 (2010) 2635–2648

2645

[53] F. Fernández, D. Borrajo, L.E. Parker, A reinforcement learning algorithm in cooperative multi-robot domains, Journal of Intelligent and Robotic Systems 43 (2005) 161–174. [54] A. Ferrein, G. Lakemeyer, Logic-based robot control in highly dynamic domains, Robotics and Autonomous Systems 56 (2008) 980–991. [55] D. Floreano, J. Urzelai, Evolutionary robots with on-line self-organization and behavioral fitness, Neural Networks 13 (2000) 431–443. [56] D. Fox, J. Ko, K. Konolige, B. Limketkai, D. Schulz, B. Stewart, Distributed multi-robot exploration and mapping, Proceedings of the IEEE 94 (2006) 1325–1339. [57] T. Fukuda, I. Takagawa, Y. Hasegawa, From intelligent robot to multi-agent robotic system, in: International Conference on Integration of Knowledge Intensive Multi-Agent Systems, 2003, pp. 413–417. [58] R.F.M. Garcia, D.J. Christensen, K. Stoy, A. Lyder, A self-reconfigurable communication network for modular robots, in: ACM First International Conference on Robot Communication and Coordination, 2007, Article 23. [59] K. Gold, B. Scassellati, Using probabilistic reasoning over time to self-recognize, Robotics and Autonomous Systems 57 (4) (2009) 384–392. [60] A. Golovinsky, M. Yim, Z. Ying, C. Eldershaw, D. Duff, PolyBot and PolyKinetics System: a modular robotic platform for education, in: Proceedings of the IEEE International Conference Robotics and Automation (ICRA’04), vol. 2, 2004, pp. 1381–1386. [61] M. Graña, I. Villaverde, J.O. Maldonado, C. Hernandez, Two lattice computing approaches for the unsupervised segmentation of hyperspectral images, Neurocomputing 72 (10–12) (2009) 2111–2120. [62] Q. Guo, M. Zhang, A novel approach for multi-agent-based intelligent manufacturing system, Information Sciences 179 (18) (2009) 3079–3090. [63] A. Grushin, J.A. Reggia, Automated design of distributed control rules for the self-assembly of prespecified artificial structures, Robotics and Autonomous Systems 56 (2008) 334–359. [64] M. Hafez, M.D. Lichter, S. Dubowsky, Optimized binary modular reconfigurable robotic devices, IEEE/ASME Transactions on Mechatronics 8 (2003) 18–25. [65] I. Harmati, K. Skrzypczyk, Robot team coordination for target tracking using fuzzy logic controller in game theoretic framework, Robotics and Autonomous Systems 57 (1) (2009) 75–86. [66] M. Hasanuzzaman, T. Zhang, V. Ampornaramveth, H. Gotoda, Y. Shirai, H. Ueno, Adaptive visual gesture recognition for human–robot interaction using a knowledge-based software platform, Robotics and Autonomous Systems 55 (2007) 643–657. [67] M. Hashimoto, K. Takahashi, Y. Matsui, Moving-object tracking with multi-laser range sensors for mobile robot navigation, in: IEEE International Conference on Robotics and Biomimetics (ROBIO 2007), 2007, pp. 399–404. [68] N. Hazon, G.A. Kaminka, On redundancy, efficiency, and robustness in coverage for multiple robots, Robotics and Autonomous Systems 56 (2008) 1102–1114. [69] N. Homayoun, G. Andrew, Real-time motion planning of an autonomous mobile manipulator using a fuzzy adaptive Kalman filter, Robotics and Autonomous Systems 55 (2) (2007) 96–106. [70] A. Howard, Multi-robot simultaneous localization and mapping using particle filters, International Journal of Robotics Research 25 (2006) 1243–1256. [71] T.-R. Hsiang, E.M. Arkin, M.A. Bender, S.P. Fekete, J.S.B. Mitchell, Algorithms for rapidly dispersing robot swarms in unknown environments, in: J.D. Boissonnat et al. (Eds.), Algorithmic Foundations of Robotics V, Springer-Verlag, 2003, pp. 77–94. [72] Y. Ikemoto, Y. Hasegawa, T. Fukuda, K. Matsuda, Gradual spatial pattern formation of homogeneous robot group, Information Sciences 171 (4) (2005) 431–445. [73] A.J. Ijspeert, Central pattern generators for locomotion control in animals and robots: a review, Neural Networks 21 (2008) 642–653. [74] P. Jan’t Hoen, K. Tuyls, L. Panait, S. Luke, J.A. La Poutré, An overview of cooperative and competitive multiagent learning, in: Learning and Adaption in Multi-Agent Systems, LNCS, vol. 3898, 2006, pp. 1–46. [75] K.G. Jolly, K.P. Ravindran, R. Vijayakumar, R. Sreerama Kumar, Intelligent decision making in multi-agent robot soccer system through compounded artificial neural networks, Robotics and Autonomous Systems 55 (2007) 589–596. [76] B.A. Jones, Kinematics and implementation of continuum manipulators, Ph.D. Thesis, Clemson University, 2005. [77] K. Jungmin, K. Yountae, K. Sungshin, An accurate localization for mobile robot using extended Kalman filter and sensor fusion, in: IEEE International Joint Conference on Neural Networks (IJCNN 2008), 2008, pp. 2928–2933. [78] A. Kamimura, H. Kurokawa, E. Yoshida, K. Tomita, S. Kokaji, S. Murata, Distributed adaptive locomotion by a modular robotic system, M-TRAN II, in: IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2004), vol. 3, 2004, pp. 2370–2377. [79] H. Karbasi, J.P. Huissoon, A. Khajepour, Uni-drive modular robots: theory, design, and experiments, Mechanism and Machine Theory 39 (2004) 183– 200. [80] A.M. Khamis, M.S. Kamel, M.A. Salichs, Cooperation: concepts and general typology, in: IEEE International Conference on Systems, Man and Cybernetics, vol. 2, 2006, pp. 1499–1505. [81] C.A. Kitts, M. Egerstedt, Design, control, and applications of real-world multirobot systems, IEEE Robotics & Automation Magazine 15 (2008) 8. [82] A. Ko, T.L. Lau, H.Y.K. Lau, Topological representation and analysis method for multi-port and multi-orientation docking modular robots, in: IEEE International Conference on Robotics and Automation (ICRA’04), vol. 3, 2004, pp. 2210–2215. [83] J.R. Kok, M.T.J. Spaan, N. Vlassis, Non-communicative multi-robot coordination in dynamic environments, Robotics and Autonomous Systems 50 (2-3) (2005) 99–114. [84] H. Kress-Gazit, D.C. Conner, H. Choset, A.A. Rizzi, G.J. Pappas, Courteous cars, IEEE Robotics & Automation Magazine 15 (2008) 30–38. [85] K.N. Krishnanand, D. Ghose, Formations of minimalist mobile robots using local-templates and spatially distributed interactions, Robotics and Autonomous Systems 53 (2005) 194–213. [86] S. Kristensen, Sensor planning with Bayesian decision theory, Robotics and Autonomous Systems 19 (1997) 273–286. [87] V. Kumar, F. Sahin, Cognitive maps in swarm robots for the mine detection application, in: IEEE International Conference on Systems, Man and Cybernetics, vol. 4, 2003, pp. 3364–3369. [88] H. Kurokawa, A. Kamimura, E. Yoshida, K. Tomita, S. Murata, S. Kokaji, Self-reconfigurable modular robot (M-TRAN) and its motion design, in: 7th International Conference on Control, Automation, Robotics and Vision (ICARCV 2002), vol. 1, 2002, pp. 2–5. [89] H. Kurokawa, E. Yoshida, K. Tomita, A. Kamimura, S. Murata, S. Kokaji, Self-reconfigurable M-TRAN structures and walker generation, Robotics and Autonomous Systems 54 (2006) 142–149. [90] M. Lee, Evolution of behaviors in autonomous robot using artificial neural network and genetic algorithm, Information Sciences 155 (1-2) (2003) 43– 60. [91] J. Lemay, L. Notash, Configuration engine for architecture planning of modular parallel robots, Mechanism and Machine Theory 39 (2004) 101–117. [92] K. Lerman, A. Martinoli, A. Galstyan, A review of probabilistic macroscopic models for swarm robotic systems, LNCS 3342 (2005) 143–152. [93] Z. Liping, M. Shugen, L. Bin, Z. Zheng, D. Zaili, C. Binggang, Position-sensing based a new docking system of RPRS, in: Control, Automation, Robotics and Vision Conference (ICARCV 2004), vol. 1, 2004, pp. 278–282. [94] I. Lopez-Juarez, M. Howarth, Knowledge acquisition and learning in unstructured robotic assembly environments, Information Sciences 145 (1–2) (2002) 89–111. [95] Z. Lu, S. Ma, B. Li, Y. Wang, Serpentine locomotion of a snake-like robot controlled by cyclic inhibitory CPG model, in: International Conference on Intelligent Robots and Systems (IROS 2005), 2005, pp. 96–101. [96] R. Lundh, L. Karlsson, A. Saffiotti, Autonomous functional configuration of a network robot system, Robotics and Autonomous Systems 56 (2008) 819– 830. [97] A. Makarenko, H. Durrant-Whyte, Decentralized Bayesian algorithms for active sensor networks, Information Fusion 7 (4) (2006) 418–433. [98] P. Manoonpong, F. Pasemann, F. Worgotter, Sensor-driven neural control for omnidirectional locomotion and versatile reactive behaviors of walking machines, Robotics and Autonomous Systems 56 (3) (2008) 265–288.

2646

R.J. Duro et al. / Information Sciences 180 (2010) 2635–2648

[99] D. Maravall, J. de Lope, J.A.H. Martin, Hybridizing evolutionary computation and reinforcement learning for the design of almost universal controllers for autonomous robots, Neurocomputing 72 (2009) 887–894. [100] J.A. Marshall, T. Fung, M.E. Broucke, G.M.T. D’Eleuterio, B.A. Francis, Experiments in multirobot coordination, Robotics and Autonomous Systems 54 (2006) 265–275. [101] F. Martin, P. Matellan, P. Barrera, J. Cañas, Localization of legged robots combining a fuzzy-Markov method and a population of extended Kalman filters, Robotics and Autonomous Systems 55 (12) (2007) 870–880. [102] A. Martinoli, K. Easton, Modeling swarm robotic systems, in: B. Siciliano, P. Dario (Eds.), Experimental Robotics VIII, Springer-Verlag, 2003, pp. 297– 306. [103] M.J. Mataric, Reinforcement learning in multi-robot domain, Autonomous Robots 4 (1997) 73–83. [104] G.M. Mathews, H. Durrant-Whyte, M. Prokopenko, Decentralised decision making in heterogeneous teams using anonymous optimisation, Robotics and Autonomous Systems 57 (3) (2009) 310–320. [105] T. Matsuo, K. Ishii, Neural oscillator based motion control system for snake-like robot, in: Symposium on Underwater Technology and Workshop on Scientific Use of Submarine Cables and Related Technologies, 2007, pp. 397–402. [106] M. McPartland, S. Nolfi, H.A. Abbass, Emergence of communication in competitive multi-agent systems: a pareto multi-objective approach, in: Proceedings of the 2005 Conference on Genetic and Evolutionary Computation, 2005, pp. 51–58. [107] F. Mondada, M. Bonani, A. Guignard, S. Magnenat, C. Studer, D. Floreano, Superlinear physical performances in a SWARM-BOT, LNCS 3630 (2005) 282– 291. [108] F. Mondada, G.C. Pettinaro, A. Guignard, I.W. Kwee, D. Floreano, J.-L. Deneubourg, S. Nolfi, L.M. Gambardella, M. Dorigo, Swarm-Bot: a new distributed robotic concept, Autonomous Robots 17 (2004) 193–221. [109] S. Murata, E. Yoshida, A. Kamimura, H. Kurokawa, K. Tomita, S. Kokaji, M-TRAN: self-reconfigurable modular robotic system, IEEE/ASME Transactions on Mechatronics 7 (2002) 431–441. [110] S. Murata, E. Yoshida, H. Kurokawa, K. Tomita, S. Kokaji, Concept of self-reconfigurable modular robotic system, Artificial Intelligence in Engineering 15 (2001) 383–387. [111] A.L. Nelson, G.J. Barlow, L. Doitsidis, Fitness functions in evolutionary robotics: a survey and analysis, Robotics and Autonomous Systems 57 (4) (2009) 345–370. [112] S. Nolfi, Evolving non-trivial behaviors on real robots: a garbage collecting robot, Robotics and Autonomous Systems 22 (1997) 187–198. [113] A. Nüchter, J. Hertzberg, Towards semantic maps for mobile robots, Robotics and Autonomous Systems 56 (2008) 915–926. [114] D. Obradovic, H. Lenz, M. Schupfner, Fusion of sensor data in Siemens car navigation system, IEEE Transactions on Vehicular Technology 56 (1) (2007) 43–50. [115] B.J. Oommen, I. Reichstein, On the problem of multiple mobile robots cluttering a workspace, Information Sciences 63 (1–2) (1992) 55–85. [116] D.J. Pack, B.E. Mullins, Toward finding an universal search algorithm for swarm robots, in: IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2003), 2003, pp. 1945–1950. [117] S. Park, S. Kim, M. Park, S.-K. Park, Vision-based global localization for mobile robots with hybrid maps of objects and spatial layouts, Information Sciences 179 (24) (2009) 4174–4198. [118] G.B. Parker, Co-evolving model parameters for anytime learning in evolutionary robotics, Robotics and Autonomous Systems 33 (2000) 13–30. [119] C.D. Pathiranage, L. Udawatta, K. Watanabe, K. Izumi, A fuzzy logic based approach to the SLAM problem using pseudolinear models with two sensors data association, in: Third International Conference on Information and Automation for Sustainability (ICIAFS 2007), 2007, pp. 70–75. [120] C.D. Pathiranage, K. Watanabe, K. Izumi, A solution to the SLAM problem based on fuzzy Kalman filter using pseudolinear measurement model, in: Annual Conference SICE, 2007, pp. 2364–2371. [121] S.A. Patterson, K.A. Knowles, Jr., B.E. Bishop, Toward magnetically-coupled reconfigurable modular robots, in: IEEE International Conference on Robotics and Automation (ICRA’04), vol. 3, pp. 2900–2905. [122] J. Pavon, J. Gomez-Sanz, A. Fernandez-Caballero, J.N.J. Valencia-Jimenez, Development of intelligent multisensor surveillance systems with agents, Robotics and Autonomous Systems 55 (2007) 892–903. [123] K. Payne, B. Salemi, P. Will, S. Wei-Min, Sensor-based distributed control for chain-typed self-reconfiguration, in: IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2004), 2004, pp. 2074–2080. [124] D. Payton, R. Estkowski, M. Howard, Compound behaviors in pheromone robotics, Robotics and Autonomous Systems 44 (2003) 229–240. [125] D. Payton, R. Estkowski, M. Howard, Pheromone robotics and the logic of virtual pheromones, LNCS 3342 (2005) 45–57. [126] D. Peleg, Distributed coordination algorithms for mobile robot swarms: new directions and challenges, LNCS 3741 (2005) 1–12. [127] G.A.S. Pereira, V. Kumar, M.F.M. Campos, Closed loop motion planning of cooperating mobile robots using graph connectivity, Robotics and Autonomous Systems 56 (2008) 373–384. [128] A. Prieto, F. Bellas, R.J. Duro, Adaptively coordinating heterogeneous robot teams through asynchronous situated coevolution, in: ICONIP (2), LNCS, vol. 5864, 2009, pp. 75–82. [129] A.H. Purnamadjaja, R.A. Russell, Pheromone communication: implementation of necrophoric bee behaviour in a robot swarm, in: IEEE Conference on Robotics, Automation and Mechatronics, vol. 2, 2004, pp. 638–643. [130] O. Purwin, R. D’Andrea, J.-W. Lee, Theory and implementation of path planning by negotiation for decentralized agents, Robotics and Autonomous Systems 56 (2008) 422–436. [131] G. Reina, A. Vargas, K. Nagatani, K. Yoshida, Adaptive Kalman filtering for GPS-based mobile robot localization, in: IEEE International Workshop on Security and Rescue Robotics (SSRR 2007), 2007, pp. 1–6. [132] W. Ren, R.W. Beard, Distributed Consensus in Multi-vehicle Cooperative Control, Springer-Verlag, 2008. [133] W. Ren, N. Sorensen, Distributed coordination architecture for multi-robot formation control, Robotics and Autonomous Systems 56 (2008) 324–333. [134] G. Resconi, A.J. van der Wal, Morphogenic neural networks encode abstract rules by data, Information Sciences 142 (1–4) (2002) 249–273. [135] R. Rocha, J. Dias, A. Carvalho, Cooperative multi-robot systems: a study of vision-based 3-D mapping using information theory, Robotics and Autonomous Systems 53 (2005) 282–311. [136] C. Ronald, Eric, Cooperative transport by ants and robots, Robotics and Autonomous Systems 30 (2000) 85–101. [137] E.M.A. Ronald, M. Sipper, Surprise versus unsurprise: implications of emergence in robotics, Robotics and Autonomous Systems 37 (2001) 19–24. [138] J.A. Rothermich, M.I. Ecemis, P. Gaudiano, Distributed localization and mapping with a robotic swarm, LNCS 3342 (2005) 58–69. [139] S. Rutishauser, N. Correll, A. Martinoli, Collaborative coverage using a swarm of networked miniature robots, Robotics and Autonomous Systems 57 (5) (2009) 517–525. [140] J.S.S. Yamada, Adaptive action selection without explicit communication for multirobot box pushing, IEEE Transactions on SMC, Part C 31 (2001) 398– 404. [141] M. Sack, N. Elkmann, T. Felsch, T. Bohme, Intelligent control of modular kinematics – the robot platform STRIUS, in: Proceedings of the 2002 IEEE International Symposium on Intelligent Control, 2002, pp. 549–553. [142] E. Sahin, Swarm robotics: from sources of inspiration to domains of application, LNCS 3342 (2005) 10–20. [143] E. Sahin, T.H. Labella, V. Trianni, J.L. Deneubourg, P. Rasse, D. Floreano, L. Gambardella, F. Mondada, S. Nolfi, M. Dorigo, SWARM-BOT: pattern formation in a swarm of self-assembling mobile robots, in: IEEE International Conference on Systems, Man and Cybernetics, 2002, pp. 6–9. [144] S. Saidani, Self-reconfigurable robots topodynamic, in: Proceedings of the IEEE International Conference on Robotics and Automation (ICRA’04), 2004, pp. 2883–2887. [145] B. Salemi, P. Will, W.M. Shen, Distributed task negotiation in self-reconfigurable robots, in: IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2003), 2003, pp. 2448–2453.

R.J. Duro et al. / Information Sciences 180 (2010) 2635–2648

2647

[146] N. Santoro, Distributed computations by autonomous mobile robots, LNCS 2234 (2001) 110–115. [147] J. Santos, R.J. Duro, J.A. Becerra, J.L. Crespo, F. Bellas, Considerations in the application of evolution to the generation of robot controllers, Information Sciences 133 (3–4) (2001) 127–148. [148] S. Sariel, T. Balch, N. Erdogan, Naval mine countermeasure missions, IEEE Robotics & Automation Magazine 15 (2008) 45–52. [149] M.F. Selekwa, D.D. Dunlap, D. Shi, E.G. Collins Jr., Robot navigation in very cluttered environments by preference-based fuzzy behaviors, Robotics and Autonomous Systems 56 (2008) 231–246. [150] K. Seong-Ho, M.W. Pryor, D. Tesar, Kinematic model and metrology system for modular robot calibration, in: Proceedings of the IEEE International Conference on Robotics and Automation (ICRA’04), 2004, pp. 2894–2899. [151] S. Sfakiotakis, D.P. Tsakiris, Biomimetic centering for undulatory robots, International Journal of Robotics Research 26 (2007) 267–1282. [152] F. Shen, O. Hasegawa, A fast nearest neighbor classifier based on self-organizing incremental neural network, Neural Networks 21 (2008) 1537–1547. [153] W.-M. Shen, P. Will, A. Galstyan, C.-M. Chuong, Hormone-inspired self-organization and distributed control of robotic swarms, Autonomous Robots 17 (2004) 93–105. [154] W. Sheng, Q. Yang, J. Tan, N. Xi, Distributed multi-robot coordination in area exploration, Robotics and Autonomous Systems 54 (2006) 945–955. [155] W. Shenga, Q. Yangb, J. Tanc, N. Xid, Distributed multi-robot coordination in area exploration, Robotics and Autonomous Systems 54 (2006) 945–955. [156] W.M. Spears, D.F. Spears, J.C. Hamann, R. Heil, Distributed, physics-based control of swarms of vehicles, Autonomous Robots 17 (2004) 137–162. [157] A. Sproewitz, R. Moeckel, J. Maye, A.J. Ijspeert, Learning to move in modular robots using central pattern generators and online optimization, International Journal of Robotics Research 27 (3-4) (2008) 423–443. [158] D.J. Stilwell, B.E. Bishop, A framework for decentralized control of autonomous vehicles, in: IEEE International Conference on Robotics and Automation (ICRA’00), 2000, pp. 2358–2363. [159] D.P. Stormont, Autonomous rescue robot swarms for first responders, in: IEEE International Conference Computational Intelligence for Homeland Security and Personal Safety (CIHSPS 2005), 2005, pp. 151–157. [160] D.P. Stormont, A. Bhatt, B. Boldt, S. Skousen, M.D. Berkemeier, Building better swarms through competition: lessons learned from the AAAI/RoboCup Rescue Robot competition, in: IROS 2003, 2003, pp. 2870–2875. [161] K. Stoy, Using cellular automata and gradients to control self-reconfiguration, Robotics and Autonomous Systems 54 (2006) 135–141. [162] A.W. Stroupe, T. Balch, Value-based action selection for observation with robot teams using probabilistic techniques, Robotics and Autonomous Systems 50 (2005) 85–97. [163] F. Stulp, M. Beetz, Combining declarative, procedural, predictive knowledge to generate, execute, and optimize robot plans, Robotics and Autonomous Systems 56 (2008) 967–979. [164] K. Sugawara, T. Watanabe, Swarming robots – collective behavior of interacting robots, in: Proceedings of the 41st SICE Annual Conference (SICE 2002), 2002, pp. 3214–3215. [165] K. Sugawara, T. Watanabe, Swarming robots-foraging behavior of simple multirobot system, in: IEEE/RSJ International Conference on Intelligent Robots and System (IROS 2002), 2002, pp. 2702–2707. [166] H. Szu, P. Chanyagorn, W. Hwang, M. Paulin, T. Yamakawa, Collective and distributive swarm intelligence: evolutional biological survey, International Congress Series, vol. 1269, Elsevier, 2004, pp. 46–49. [167] Y. Takahashi, K. Edazawa, K. Noma, M. Asada, Simultaneous learning to acquire competitive behaviors in multi-agent system based on a modular learning system, in: IEEE/RSJ Intelligent Robots and Systems (IROS 2005), 2005, pp. 2016–2022. [168] H.G. Tanner, D.K. Christodoulakis, Decentralized cooperative control of heterogeneous vehicle groups, Robotics and Autonomous Systems 55 (2007) 811–823. [169] T. Tao, Y. Huang, J. Yuan, F. Sun, Multi-robot cooperative map building in unknown environment considering estimation uncertainty, in: Control and Decision Conference (CCDC 2008), 2008, pp. 2896–2901. [170] Y. Terada, S. Murata, Automatic assembly system for a large-scale modular structure – hardware design of module and assembler robot, in: Intelligent Robots and Systems (IROS 2004), 2004, pp. 2349–2355. [171] H. Tokashiki, H. Amagai, S. Endo, K. Yamada, J. Kelly, Development of a transformable mobile robot composed of homogeneous gear-type units, in: Intelligent Robots and Systems (IROS 2003), 2003, pp. 1602–1607. [172] B. Tovar, L. Muñoz-Gomez, R. Murrieta-Cid, M. Alencastre-Miranda, R. Monroy, S. Hutchinson, Planning exploration strategies for simultaneous localization and mapping, Robotics and Autonomous Systems 54 (2006) 314–331. [173] V. Trianni, S. Nolfi, M. Dorigo, Cooperative hole avoidance in a swarm-bot, Robotics and Autonomous Systems 54 (2006) 97–103. [174] P. Vadakkepat, X. Peng, B.K. Quek, T.H. Lee, Evolution of fuzzy behaviors for multi-robotic systems, Robotics and Autonomous Systems 55 (2007) 146– 161. [175] S. Vassilvitskii, J. Kubica, E. Rieffel, J. Suh, M. Yim, On the general reconfiguration problem for expanding cube style modular robots, in: IEEE International Conference on Robotics and Automation (ICRA’02), 2002, pp. 801–808. [176] I. Villaverde, A. d’Anjou, M. Graña, Morphological neural networks and vision based simultaneous localization and mapping, Integrated ComputerAided Engineering 14 (2007) 355–363. [177] J. Wessnitzer, C. Melhuish, Collective decision-making and behaviour transitions in distributed ad hoc wireless networks of mobile robots: targethunting, LNCS 2801 (2003) 893–902. [178] E. Yang, D. Gu, Multiagent reinforcement learning for multi-robot systems: a survey, Tech. Report, Univ. Essex, 2004. [179] G. Yang, I.M. Chen, A novel kinematic calibration algorithm for reconfigurable robotic systems, in: IEEE International Conference on Robotics and Automation, 1997, pp. 3197–3202. [180] G. Yang, I.M. Chen, Task-based optimization of modular robot configurations: minimized degree-of-freedom approach, Mechanism and Machine Theory 35 (2000) 517–540. [181] F. Yangiong, Z. Xifang, W.L. Xu, Kinematics and dynamics of reconfigurable modular robots, in: IEEE International Conference on Systems, Man, and Cybernetics, 1998, pp. 3325–3334. [182] M. Yim, Locomotion with a unit-modular reconfigurable robot, Technical Report, CS-TR-95-1536, Department of Computer Science: Stanford University, 1995. [183] M. Yim, D.G. Duff, K.D. Roufas, PolyBot: a modular reconfigurable robot, in: Proceedings of the IEEE International Conference on Robotics and Automation (ICRA’00), 2000, pp. 514–520. [184] M. Yim, Z. Ying, D. Duff, Modular robots, IEEE Spectrum 39 (2002) 30–34. [185] M. Yim, Z. Ying, K. Roufas, D. Duff, C. Eldershaw, Connecting and disconnecting for chain self-reconfiguration with PolyBot, IEEE/ASME Transactions on Mechatronics 7 (2002) 442–451. [186] T. Yong Chye, B.E. Bishop, Evaluation of robot swarm control methods for underwater mine countermeasures, in: Proceedings of the Thirty-Sixth Southeastern Symposium on System Theory, 2004, pp. 294–298. [187] E. Yoshida, S. Murata, A. Kamimura, K. Tomita, H. Kurokawa, S. Kokaji, Self-reconfigurable modular robots – hardware and software development in AIST, in: IEEE International Conference on Robotics, Intelligent Systems and Signal Processing, 2003, pp. 339–346. [188] E. Yoshida, S. Murata, A. Kamimura, K. Tomita, H. Kurokawa, S. Kokaji, Evolutionary synthesis of dynamic motion and reconfiguration process for a modular robot M-TRAN, in: Proceedings of the IEEE International Symposium on Computational Intelligence in Robotics and Automation, vol. 2, 16– 20 July 2003, 2003, pp. 1004–1010. [189] A. Zaerpoora, M.N. Ahmadabadia, M.R. Barunia, Z.D. Wang, Distributed object transportation on a desired path based on constrain and move strategy, Robotics and Autonomous Systems 50 (2005) 115–128.

2648

R.J. Duro et al. / Information Sciences 180 (2010) 2635–2648

[190] D. Zhang, G. Xie, J. Yu, L. Wang, Adaptive task assignment for multiple mobile robots via swarm intelligence approach, Robotics and Autonomous Systems 55 (2007) 572–588. [191] D.G. Zhang, K.Y. Zhu, L. Lan, Analysis of neural oscillator for bio-inspired robot control, in: IEEE Conference on Robotics, Automation and Mechatronics, 2006, pp. 1–6. [192] W.J. Zhang, S.N. Liu, Q. Li, Data/knowledge representation of modular robot and its working environment, Robotics and Computer-Integrated Manufacturing 16 (2000) 143–159. [193] J. Zhao, J. Jiang, X. Zang, Cooperative multi-robot map-building based on genetic algorithms, in: Proceedings of the IEEE International Conferecne on Information Acquisition, 2006, pp. 360–364. [194] T. Zheng, L. Yang, Optimal ant colony algorithm based multi-robot task allocation and processing sequence scheduling, in: 7th World Congress on Intelligent Control and Automation (WCICA 2008), 2008, pp. 5693–5698. [195] C. Zhou, Robot learning with GA-based fuzzy reinforcement learning agents, Information Sciences 145 (1-2) (2002) 45–68. [196] Y. Zou, D. Luo, A modified ant colony algorithm used for multi-robot odor source localization, LNCS 5227 (2008) 502–509.