Semi-heterarchical architecture to AGV adjustable autonomy within FMSs

13th IFAC Workshop on Intelligent Manufacturing Systems 13th IFAC Workshop on Manufacturing Systems Available online at www.sciencedirect.com 13th IFAC Workshop on Intelligent Intelligent Manufacturing Systems August 12-14, 2019. Oshawa, Canada 13th IFAC Workshop on Intelligent Manufacturing Systems August 12-14, 2019. Canada August 12-14, 2019. Oshawa, Oshawa, Canada 13th IFAC Workshop on Intelligent Manufacturing Systems August 12-14, 2019. Oshawa, Canada August 12-14, 2019. Oshawa, Canada

ScienceDirect

IFAC PapersOnLine 52-10 (2019) 7–12

Semi-heterarchical architecture to AGV adjustable autonomy within FMSs Semi-heterarchical architecture to AGV adjustable autonomy within FMSs Semi-heterarchical architecture to AGV adjustable autonomy within FMSs Semi-heterarchical architecture to AGV adjustable autonomy within FMSs Sergio Ramiro Gonzalez*. Gabriel Mauricio Zambrano** Ivan Fernando Mondragon*** Sergio Ramiro Gonzalez*. Gabriel Mauricio Zambrano** Ivan Fernando Mondragon*** Semi-heterarchical architecture to AGV adjustable autonomy within FMSs Sergio Ramiro Gonzalez*. Gabriel Mauricio Zambrano** Ivan Fernando Mondragon*** Sergio Ramiro Gonzalez*. Gabriel Mauricio Zambrano** Ivan Fernando Mondragon***

Sergio Ramiro Javeriana Gonzalez*. Gabriel Mauricio Zambrano** Ivan Fernando Mondragon*** ** Pontificia Universidad (e-mail: [email protected] ). **Pontificia Universidad Javeriana (e-mail: Universidad Javeriana (e-mail: [email protected] ). **Pontificia Universidad Javeriana (e-mail: ** Pontificia Pontificia Universidad Javeriana (e-mail: [email protected] ). **Pontificia Universidad Javeriana (e-mail: [email protected] ). *** Pontificia Universidad Javeriana (e-mail: [email protected] ). Pontificia Universidad Javeriana (e-mail: [email protected] ). **Pontificia Universidad Javeriana (e-mail: [email protected] ). *** Pontificia Universidad Javeriana (e-mail: [email protected] ). [email protected] ). *** Pontificia Universidad Javeriana (e-mail: [email protected] ). * Pontificia Universidad Javeriana (e-mail: [email protected] ). **Pontificia Universidad Javeriana [email protected]). *** Pontificia Universidad Javeriana (e-mail: [email protected]).(e-mail: [email protected]). *** Pontificia Universidad Javeriana (e-mail: [email protected]). Abstract: A flexible system (FMS) is aa highly manufacturing system in which Abstract: A flexible manufacturing manufacturing system (FMS) is highly integrated integrated manufacturing system in which Abstract: A manufacturing system (FMS) is integrated manufacturing system in there is some amount of flexibility that allows the system to react in case of changes, whether Abstract: A flexible flexible manufacturing system (FMS) is aa highly highly integrated manufacturing system predicted in which which there is some amount of flexibility that allows the system to react in case of changes, whether predicted there is some amount of flexibility that allows the system to react in case of changes, whether predicted Abstract: A flexible manufacturing system (FMS) is a highly integrated manufacturing system in which or unpredicted. Automated guided vehicles (AGVs) for FMSs because they provide there is some amount of flexibility that allows the systemare to suitable react in case of changes, whether predicted or unpredicted. Automated guided vehicles (AGVs) are suitable for FMSs because they provide or unpredicted. Automated guided vehicles (AGVs) are suitable for FMSs because they provide there is some amount of flexibility that allows the system to react in case of changes, whether predicted flexibility, adjustability and the connection processing by handling raw materials, subor unpredicted. Automated guided vehiclesof (AGVs) aresubsystems suitable for FMSs because they provide flexibility, adjustability and the connection of processing by handling raw materials, subflexibility, adjustability and the connection of processing subsystems by handling raw materials, subor unpredicted. Automated guided vehicles (AGVs) aresubsystems suitable for FMSs because they provide assemblies or finished products. The static level of autonomy granted to AGVs affects their flexibility in flexibility, adjustability and the connection of processing subsystems by handling raw materials, subassemblies or finished products. The static level of autonomy granted to AGVs affects their flexibility in assemblies or finished products. The static level of autonomy granted to AGVs affects their flexibility in flexibility, adjustability and the connection of processing subsystems by handling raw materials, subdealing with perturbations, efficiency and the contribution to global performance. This paper presents assemblies or finished products. The static level of autonomy granted to AGVs affects their flexibility inaa dealing with perturbations, efficiency and the contribution to global performance. This paper presents dealing with perturbations, efficiency and the contribution to global performance. This paper presents assemblies or finished products. The static level of autonomy granted to AGVs affects their flexibility inaa semi-heterarchical architecture to AGVs’ autonomy control to mitigate perturbations of FMS and dealing with perturbations, efficiency and the contribution to global performance. This paper presents semi-heterarchical architecture to AGVs’ autonomy control to mitigate perturbations of FMS and semi-heterarchical architecture to AGVs’ autonomy control to mitigate perturbations of FMS and dealing with perturbations, efficiency and the contribution to global performance. This paper presents increase their overall performance. approach is based on the architecture between semi-heterarchical architecture to This AGVs’ autonomy control tosemi-heterarchical mitigate perturbations of FMS anda increase their overall performance. This approach is based on the semi-heterarchical architecture between increase their overall performance. This approach is based on the semi-heterarchical architecture between semi-heterarchical architecture to AGVs’ autonomy control to mitigate perturbations of FMS and AGVs belief-desired-intention model decision-making under normal and disturbance increaseusing their overall performance. ThisBDI approach isfor based on the semi-heterarchical architecture between AGVs using belief-desired-intention BDI model for decision-making under normal and disturbance AGVs using belief-desired-intention BDI model for decision-making under normal and disturbance increase their overall performance. This approach is based on the semi-heterarchical architecture between scenarios. The effectiveness of the proposed approach is demonstrated via a case study. We conclude that AGVs using belief-desired-intention BDI model forisdecision-making under normal andconclude disturbance scenarios. The effectiveness of the proposed approach demonstrated via aa case study. We that scenarios. The effectiveness of the proposed approach demonstrated via case study. We that AGVs using belief-desired-intention BDI model foris decision-making under normal andconclude disturbance adjustable autonomy results in better performance than the classic static version. scenarios. The effectiveness of the proposed approach is demonstrated via a case study. We conclude that adjustable autonomy results in better performance than the classic static version. adjustable autonomy results in better performance than the classic static version. scenarios. The effectiveness of the proposed approach is demonstrated via a case study. We conclude that adjustable autonomy results in better performance than the classic static version. Keywords: Autonomy, adjustable autonomy, automated guided vehicle (AGV), semi-heterarchical © 2019, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. Allsemi-heterarchical rights reserved. adjustable autonomy results in better performance than the classic static version. Keywords: Autonomy, adjustable autonomy, automated guided vehicle (AGV), Keywords: adjustable autonomy, automated guided (AGV), semi-heterarchical and flexible manufacturing system (FMS). Keywords: Autonomy, Autonomy, adjustable autonomy, automated guided vehicle vehicle (AGV), semi-heterarchical architecture, cyber-physical systems (CPS), (CPS), and flexible manufacturing system (FMS). architecture, cyber-physical systems and flexible manufacturing system (FMS). architecture, cyber-physical systems (CPS), Keywords: Autonomy, adjustable autonomy, automated guided vehicle (AGV), architecture, cyber-physical systems (CPS), and flexible manufacturing system (FMS). semi-heterarchical  manufacturing system (FMS). architecture, cyber-physical systems (CPS), and flexible  provide high flexibility and interaction between processing  provide high flexibility and interaction between processing 11 INTRODUCTION provide high flexibility and interaction between processing  INTRODUCTION subsystems in CPS. To deal with disturbances, the AGV provide high flexibility and interaction between processing 11 INTRODUCTION subsystems in CPS. To deal with disturbances, the AGV INTRODUCTION subsystems in CPS. To deal with disturbances, the AGV provide high flexibility and interaction between processing must be able to make quick decisions autonomously and During the last two decades, the market for industrial subsystems in CPS. To deal with disturbances, the AGV 1 INTRODUCTION During the last two decades, the market for industrial must be able to make quick decisions autonomously and must be able to make quick decisions autonomously and During the last two decades, the market for industrial subsystems in CPS. To deal with disturbances, the AGV recognize and overcome a wide range of perturbations with manufacturing companies has become more and more must be able to make quick decisions autonomously and During the last two decades, the market for industrial recognize and overcome a wide range of perturbations with manufacturing companies has become more and more recognize and overcome a wide range of perturbations with manufacturing companies has become more and more must be able to make quick decisions autonomously and During the last two decades, the market for industrial various degrees of severity (Park & Tran, 2012). Such globalized, highly competitive and presents rapidly changing recognize and overcome a wide range of perturbations with manufacturing companies has become more and more various degrees of severity (Park & Tran, 2012). Such globalized, highly competitive and presents rapidly changing various degrees of severity (Park & Tran, 2012). Such globalized, highly competitive and presents rapidly changing recognize and overcome a wide range of perturbations with manufacturing companies has become more and more autonomous behaviour is aa very desirable characteristic of customer expectations. Therefore, new requirements are various degrees of severity (Park & Tran, 2012). Such globalized, highly competitive and presents rapidly changing autonomous behaviour is desirable characteristic of customer expectations. Therefore, new requirements are autonomous behaviour is aa very very desirable characteristic of customer expectations. Therefore, new requirements are various degrees of (KUSTAK, severity (Park & Tran, 2012). Such globalized, highly competitive and presents rapidly changing advanced systems 1985). The definition imposed on the operation of flexible manufacturing systems autonomous behaviour is very desirable characteristic of customer expectations. Therefore, new requirements are advanced systems (KUSTAK, 1985). The definition of imposed on the operation of flexible manufacturing systems advanced systems (KUSTAK, 1985). The definition imposed on the operation of flexible manufacturing systems autonomous behaviour is a very desirable characteristic of customer expectations. Therefore, new requirements are autonomy is an essential concept in rational entities that (FMS)(Gabriel Zambrano Rey, 2014). Examples of such advanced systems (KUSTAK, 1985). The definition of imposed on the operation of flexible manufacturing systems autonomy is an essential concept in rational entities that (FMS)(Gabriel Zambrano Rey, 2014). Examples of such autonomy is an essential concept in rational entities that (FMS)(Gabriel 2014). Examples of such advanced systems (KUSTAK, 1985). Themaking definition of imposed on the Zambrano operation ofRey, flexible manufacturing systems support the reasoning, planning and decision required FMS requirements include the capacity to respond to any autonomy is an essential concept in rational entities that (FMS)(Gabriel Zambrano Rey, 2014). Examples of such support the isreasoning, reasoning, planning and decision decision making required FMS requirements include the capacity to respond to any support the planning and making required FMS requirements include the capacity to respond to any autonomy an essential concept in rational entities that (FMS)(Gabriel Zambrano Rey, 2014). Examples of such to achieve strategic self-directed goals. To be autonomous, a disturbance in real-time, achieve fault-tolerance and support the reasoning, planning and decision making required FMS requirements include theachieve capacityfault-tolerance to respond to any achieve self-directed goals. To be autonomous, disturbance in real-time, real-time, and to to achieve strategic self-directed goals. To be autonomous, disturbance in fault-tolerance and support thestrategic reasoning, planning and decision making requiredaaa FMS requirements include theachieve capacity to maintaining respond to any system ‘... should have the ability to learn, react, interact, hardware/software re-configurability, while the to achieve strategic self-directed goals. To be autonomous, disturbance in real-time, achieve fault-tolerance and hardware/software re-configurability, while maintaining the system ‘... should have the ability to learn, interact, system ‘... should the ability to react, interact, hardware/software re-configurability, while maintaining the achieve strategic self-directed goals. To be react, autonomous, a disturbance in performance. real-time, achieve fault-tolerance and make decisions, and gain information about their expected global These requirements can be met system ‘... should have have the ability to learn, learn, react, interact, hardware/software re-configurability, while maintaining the to make decisions, and gain information about their expected global performance. These requirements can be met make decisions, and gain information about their expected global performance. These requirements can be met system ‘... should have the ability to learn, react, interact, hardware/software re-configurability, while maintaining the environment without human or other entities’ intervention ...’ by proposing innovative control manufacturing architectures, decisions, and gain information about their expected global performance. These requirements can be met make without human or other entities’ intervention ...’ by proposing innovative control manufacturing architectures, environment without human or other entities’ intervention ...’ by proposing innovative control manufacturing architectures, make decisions, and gain information about their expected global performance. These requirements can be met environment (Dorais, Bonasso, Kortenkamp, Pell, & Schreckenghost, for instance, within the Industry 4.0 concept. environment without human or other entities’ intervention ...’ by proposing innovative control manufacturing architectures, (Dorais, Bonasso, Kortenkamp, Pell, & Schreckenghost, for instance, within the Industry 4.0 concept. (Dorais, Bonasso, Kortenkamp, Pell, & Schreckenghost, for instance, within the Industry 4.0 concept. environment without human or other entities’ intervention ...’ by proposing innovative control manufacturing architectures, 1999). (Dorais, Bonasso, Kortenkamp, Pell, & Schreckenghost, for instance, within the Industry 4.0 concept. 1999). The Industry 4.0 concept proposes smart factories based on 1999). (Dorais, Bonasso, Kortenkamp, Pell, & Schreckenghost, for instance, within the Industry 4.0 concept. The Industry 4.0 concept proposes smart factories based on 1999). The Industry concept smart factories on major challenge is to ensure that the level of autonomy is cyber-physical (CPS) that several The Industry 4.0 4.0systems concept proposes proposes smartorchestrate factories based based on A 1999). A major challenge challenge is is to to ensure ensure that that the the level level of of autonomy autonomy is is cyber-physical systems (CPS) that orchestrate several A major cyber-physical systems (CPS) that orchestrate several The Industry 4.0 concept proposes smart factories based on adjusted to meet the global performance expectations A major challenge is to ensure that the level of autonomy is technologies, such as artificial intelligence, IoT (Internet of cyber-physical systems (CPS) that orchestrate several adjusted to meet the global performance expectations technologies, such as artificial intelligence, IoT (Internet of to meet global performance technologies, such as intelligence, IoT (Internet of A major challenge isthe to ensure that theother level entities of expectations autonomy is cyber-physical systems (CPS) that orchestrate several imposed by FMS control and help to fulfil adjusted to the meet global expectations Things), big data, and cloud computing. CPS improve technologies, such as artificial artificial intelligence, IoTcan (Internet of adjusted imposed by the FMSthe control and performance help other other entities entities to fulfil fulfil Things), big data, and cloud computing. CPS can improve imposed by the FMS control and help to Things), big data, and cloud computing. CPS can improve adjusted to meet the global performance expectations technologies, such as artificial intelligence, IoT (Internet of their own local Adjustable a property of the goals. FMS control andautonomy help otheris to fulfil agility and responsiveness meet the aforementioned Things), big data, and cloud to computing. CPS can improve imposed their ownby local goals. Adjustable autonomy isentities property of aaa agility and responsiveness to meet the aforementioned their own local Adjustable autonomy is aaa property of agility and responsiveness to meet the aforementioned imposed by the goals. FMS control and help other entities to fulfil Things), big data, and cloud computing. CPS can improve decisional entity that allows the changing of the entity’s level their own local goals. Adjustable autonomy is property of a requirements. However, the complexity of computing and agility and responsiveness to meet the aforementioned decisional entitygoals. that allows allows the changing changing of isthe thea property entity’s level level requirements. However, the complexity of computing and decisional entity that the of entity’s requirements. However, the complexity of computing and their own local Adjustable autonomy of a agility and responsiveness to lot meetof the aforementioned of autonomy (LOA) from a predefined set of levels, during decisional entity that allows the changing of the entity’s level physical dynamics brings a challenges in the requirements. However, the complexity of computing and of autonomy (LOA) from aa the predefined set of levels, during physical dynamics brings lot of challenges in the of autonomy (LOA) from set levels, during physical dynamics brings lot in the decisional entity thatand allows changing of of thekey entity’s level requirements. However, theasaaacomplexity of computing system operation particularly during situations of autonomy (LOA) from a predefined predefined set of levels, during development of CPS, such interaction and integration of physical dynamics brings lot of of challenges challenges in and the system operation and particularly during key situations development of CPS, such as interaction and integration of system operation and particularly during key situations development of CPS, such as interaction and integration of of autonomy (LOA) from a predefined set of levels, during physical dynamics brings a interaction lot ofandchallenges in the (Dorais et al., 1999). The control allocates responsibilities to system operation and particularly during key situations heterogeneous processing subsystems material handling development of CPS, such as and integration of heterogeneous subsystems and material handling (Dorais et al., 1999). 1999). Theparticularly control allocates allocates responsibilities to et al., The control to heterogeneous processing subsystems and material handling system operation and duringresponsibilities key situations development ofprocessing CPS, such asCPS interaction and integration of (Dorais decisional entities, determines the relationships between them (Dorais et al., 1999). The control allocates responsibilities to systems. For that reason, the architecture must describe heterogeneous processing subsystems and material handling decisional entities, determines the relationships between them systems. For that reason, the CPS architecture must describe decisional determines the between them systems. For that reason, the CPS must describe (Dorais et entities, al., 1999). The control allocates responsibilities to heterogeneous processing subsystems and material handling and establishes the coordination mechanisms for the decisional entities, determines the relationships relationships between how the processing subsystems, including the material systems. For that reason, the CPS architecture architecture must describe and establishes the coordination mechanisms for them the how the processing subsystems, including the material and establishes the coordination mechanisms for the how the processing subsystems, including the material decisional entities, determines the relationships between them systems. For that reason, the CPS architecture must describe execution of control The existing publications on and establishes thedecisions. coordination mechanisms for the handling are interconnected, interact with other how the system, processing subsystems, including theeach material execution of control control decisions. The existing existing publications on handling are interconnected, interact with other execution of decisions. The publications on handling system, are interconnected, interact with each other and establishes thehave coordination mechanisms forbased the how the system, processing subsystems, including theeach material autonomous systems focused on decision making execution of control decisions. The existing publications on and cooperate. Additionally, the CPS architecture must handling system, are interconnected, interact with each other autonomous systems have focused on decision making based and cooperate. Additionally, the CPS architecture must have focused decision making based and cooperate. Additionally, the architecture must execution of systems control decisions. Theon existing publications on handling system, are interconnected, interact with each other on autonomy and decision-making for autonomous autonomous have focused oncontrol decision based guarantee strategic optimization. The material handling and cooperate. Additionally, the CPS CPS architecture must autonomous on autonomysystems and decision-making decision-making control formaking autonomous guarantee strategic optimization. The material handling on autonomy and control for autonomous guarantee strategic optimization. The material handling autonomous systems have focused on decision making based and cooperate. Additionally, the CPS architecture must entities (K. Barber & Martin, 1999) (Bob van der Vecht, on autonomy and decision-making system (MHS) is of the highest importance for the CPS, since guarantee strategic optimization. The material handling entities (K. Barber Barber & Martin, Martin, 1999) 1999)control (Bob for vanautonomous der Vecht, Vecht, system (MHS) is of the highest importance for the CPS, since entities (K. & (Bob van der system (MHS) is of the highest importance for the CPS, since on autonomy and &decision-making control for autonomous guarantee strategic optimization. The that material handling Dignum, Meyer, Neef, 2008) (Castelfranchi, 1994) (B entities (K. Barber & Martin, 1999) (Bob van der Vecht, the material handling can guarantee raw materials, system (MHS) is of the highest importance for the CPS, since Dignum, Meyer, & Neef, 2008) (Castelfranchi, 1994) (B the material handling can guarantee that raw materials, Dignum, Meyer, & Neef, 2008) (Castelfranchi, 1994) (B the material handling can guarantee that raw materials, entities (K. Barber & Martin, 1999) (Bob van der Vecht, system (MHS) is of the highest importance forraw the CPS, since Vecht, 2009) (Falcone & Castelfranchi, 2001), (Bob van der Dignum, Meyer, & Neef, 2008) (Castelfranchi, 1994) (B primary assemblies, and articles ready for shipping are at the the material handling can guarantee that materials, primary assemblies, and articles ready for shipping are at the Vecht, 2009) (Falcone & Castelfranchi, 2001), (Bob van der Vecht, 2009) (Falcone & 2001), (Bob van primary assemblies, and articles ready shipping are Dignum, Meyer, & Neef, 2008) (Castelfranchi, 1994) (B the material handling guarantee that rawdirectly materials, Vecht et al., 2008) (Schurr, Marecki, Lewis, Tambe, & Vecht, 2009) & Castelfranchi, Castelfranchi, van der der right place at the right time, hence affecting primary assemblies, and can articles ready for for shipping are at at the the Vecht et al., al., (Falcone 2008) (Schurr, (Schurr, Marecki,2001), Lewis,(Bob Tambe, & right place at the right time, hence affecting directly the Vecht et 2008) Marecki, Lewis, Tambe, & right place at the right time, hence affecting directly Vecht, 2009) (Falcone & Castelfranchi, 2001), (Bob van der primary assemblies, and articles ready for shipping are at the Scerri, 2005) (Maheswaran, Tambe, Varakantham, & Myers, Vecht et al., 2008) (Schurr, Marecki, Lewis, Tambe, & production performance indicator. right place at the right time, hence affecting directly the Scerri, 2005) (Maheswaran, Tambe, Varakantham, & Myers, production performance indicator. 2005) Tambe, Varakantham, & production Vecht (Pynadath, et al., (Maheswaran, 2008) (Schurr, Marecki, Lewis, Tambe, & right place performance at the rightindicator. time, hence affecting directly the Scerri, 2003) Scerri, & Tambe, 2001) (Gunderson & Scerri, 2005) (Maheswaran, & Myers, Myers, production performance indicator. 2003) (Pynadath, Scerri, & &Tambe, Tambe,Varakantham, 2001) (Gunderson (Gunderson & Automated guided vehicles (AGVs) are used for material 2003) (Pynadath, Scerri, Tambe, 2001) & Scerri, 2005) (Maheswaran, Tambe, Varakantham, & Myers, production performance indicator. Automated guided vehicles (AGVs) are used for material Martin, 1999) 2003) (Pynadath, Scerri, & Tambe, 2001) (Gunderson & Automated guided vehicles (AGVs) are used for material Martin, 1999) handling, aiming for delivering high system integration, Automated guided vehicles (AGVs) are used for material Martin, 1999) 2003) (Pynadath, Scerri, & Tambe, 2001) (Gunderson & handling, aiming delivering high system integration, handling, for delivering high integration, Automated guided for vehicles (AGVs) used for material Martin, 1999) handling, aiming aiming for delivering higharesystem system integration, Martin, 1999) 2405-8963 2019, IFAC of Automatic Control) Hosting by Elsevier Ltd. All rights reserved. handling,© aiming for(International delivering Federation high system integration,

Peer review under responsibility of International Federation of Automatic 2019 IFAC 7 Control. Copyright@ Copyright@ 2019 7 10.1016/j.ifacol.2019.10.003 Copyright@ 2019 IFAC IFAC 7 Copyright@ 2019 IFAC 7 Copyright@ 2019 IFAC 7

2019 IFAC IMS 8 August 12-14, 2019. Oshawa, Canada Sergio Ramiro Gonzalez et al. / IFAC PapersOnLine 52-10 (2019) 7–12

The LOA can be changed depending on the type of control architecture such as global (semi-heterarchical), local (heterarchical) or reconfigurable control(Hülsmann & Windt, 2007) (d’Inverno, Luck, & others, 1996) (Jensen & Kristensen, 2009). First, when the control is semiheterarchical, the control is divided into several decisional sub-activities or levels to reduce the complexity inherent in centralized approaches. Dividing the control into hierarchically dependent sub-activities with decreasing time ranges (i.e., strategic, tactic and operational, such as planning, scheduling and supervising) assigned to hierarchically dependent decisional entities allows the maintenance of sufficient optimal global control. For example, Castelfranchi and Falcone in (Falcone & Castelfranchi, 2001), and Barber and Martin in (K. Barber & Martin, 1999) presented an adaptive decision-making framework in which global entities propose strategies to the group and thereby change their own autonomy level. In this way, adjustable autonomy becomes a group process, because other entities can accept or reject proposed decision-making strategies. Nevertheless, such global autonomy control becomes more complicated with the rigidity of the control structure, which implies a weak response to change (i.e., supports less agility and reactivity), particularly when the amount of resources increases, among other factors (i.e., communication delays, complex decision making, and breakdowns) (González, Mondragón, Zambrano, Hernandez, & Montaña, 2017). Second, when the control is heterarchical, it advocates more decentralization of control decisions. The idea is to allow the decisional entities to work together so that they can react quickly instead of requesting control decisions from upper decisional levels. The last type of control, switching control, combines the two aforementioned controls. In this new control, adjustable autonomy is achieved by switching FMS control between local and global. The switching control tries to balance the overall performance with reactivity to perturbations (Leith, Shorten, Leithead, Mason, & Curran, 2003).

adaptation and to mitigate the impact of perturbations (K. S. Barber & Martin, 2000). Second, these approaches use rigid one-shot transfers of control that can result in unacceptable autonomy values (Meystel & others, 2000). Last, despite the importance of the autonomy control systems, research has often overlooked this topic, as the related literature is scarce and does not provide developers with a comprehensive framework to achieve more effective control. The first shortcomings are developed in this paper and correspond to the following question: ‘How can adjustable autonomy be achieved?’ Our hypothesis to answer this question lies in decentralized control architectures because of their benefits in term of the mitigation of perturbations. This work is focused on a CPS semi-heterarchical architecture that tackles particularly the material handling system. This approach is motivated on granting AGVs dynamic varying autonomy levels of independence. The purpose is that AGVs must perform well under significant disturbances for extended periods of time, and they must be able to compensate for system failures without external intervention to improve the FMS’s overall performance. This paper is organized as follows. Section 2 presents the approach proposed for autonomy control, describing the FMS architecture supporting the proposed adjustable autonomy control. The test method is shown in Subsection 3.2. Section 4 describes the experimental results and discussion. Finally, Section 5 presents future work and the conclusion of this paper. 2

DESCRIPTION OF THE ARCHITECTURE PROPOSED

To overcome the aforementioned shortcomings, this section describes a semi-heterarchical approach that focuses particularly on adjustable autonomy. The decisional entity (DE) carry out decision-making activities within a transport decisional process. The problem of decision making within FMSs becomes difficult, requiring the participation of all the DEs, such as processing machines (i.e., processing resources, CNC, industrial robots, etc.), AGVs, conveyor, orders and products. Architecture is a map of the internals of an entity and its data structure, the operations that may be performed on these data structures, and the control of information between these data structures (Farahvash & Boucher, 2004). The architecture defines, how is the information flow? How it is being perceived? How it is being transformed? And finally, how decisions are being made? The architecture has the functional requests of the system and aims to satisfy these requirements (Vuković & Miljković, 2009). The architecture allows vDEs to accomplish activities autonomous with a significant degree of coherence. Due to the design of the architecture, the use of the global knowledge is fortunately not detrimental to the processing requirements of the individual entities. Thus, the use of global knowledge can be incorporated into architecture without impacting the global performance. In order to accomplish the requirements described aforementioned an architecture for vDEs autonomy control is presented in the next subsection. The architecture is constituted by one approach called as semi-heterarchical.

The aforementioned studies have investigated various approaches addressing adjustable autonomy from different perspectives, but the authors have mainly focused on five ways to achieve such adjustability. First, the LOA implies satisfactory performance under significant uncertainties due to the ability to compensate for system failures without complicating the operational adaptation. Second, some strategies involve transferring control capabilities between entities or taking shorter-term decisions (Schurr et al., 2005) (Gunderson & Martin, 1999) (Scerri, Pynadath, & Tambe, 2002). Third, extensive literature has dealt with entities’ characteristics, such as static autonomy, local views and decentralization of decision making; however, relatively little attention has been directed to studying the influence between entities for designing FMS control structures based on adjustable autonomy (Hülsmann & Windt, 2007) (d’Inverno, Luck, & others, 1996) (Jensen & Kristensen, 2009). Last, the design of FMS control systems does not include the ability to adjust the level of autonomy during system execution. Unfortunately, these approaches reveal three key shortcomings. First, in some implementations, the LOA uses a simple discrete set of levels comprising all states for 8

2019 IFAC IMS August 12-14, 2019. Oshawa, Canada Sergio Ramiro Gonzalez et al. / IFAC PapersOnLine 52-10 (2019) 7–12

2.1 The semi-heterarchical approach.

9

senses changes in its environment and reacts to those changes. Its role is to guarantee good performances based on the initial state.

The semi-heterarchical approach is formed by an AGVG and a set of AGV local decisional entities (AGVLs), such that AGVL = {AGVL1 , AGVL2 , … AGVLJ }, as seen in Figure 1. Each AGVLj is assigned to control one AGV. The AGVG follows a global objective function (GOF), as depicted in Figure 1. The semi-heterarchical approach has been used to model the interactions among an AGVG and AGVLs. This approach is characterized by the existence of several control levels (e.g., master/slave relationship). Since, the approach proposed is constituted for two control levels called as global decisional level (GDL) and local decisional level (LDL) that form a pyramid. The upper-level GDL is conformed for an AGVG. The LDL is conformed for a set of AGVLs. The lowlevel performs the commands of the level above it by controlling each AGVL. The semi-heterarchical approach utilizes the global goals of the FMS system and/or global knowledge about the AGVG current or upcoming actions to direct individual AGVL actions. With these control, an AGVG is able to influence actions toward AGVL goals that cannot be sensed in its own local environment. To better understand the implications of the use of global control, let us look individually at the two types of information utilized by this control: global goals and global knowledge. The first type is global goals of an AGVG that indicate the overall actions that the transport system is required to accomplish. These goals are typically imposed upon the AGVG by machine global decisional entity (sGDE) or product global decisional entity (pGDE)(Gabriel Zambrano Rey, 2014) (Jimenez, Bekrar, Trentesaux, & Leitão, 2016).

Figure 1 Control loop behaviour of AGVL autonomy.

In this approach, the AGVG becomes capable of processing more information than AGVLs. Communication exchanges in this architecture are restricted to interactions between AGVG to AGVLs and AGVG to machine global decisional entity (sGDE), and AGVG to product global decisional entity (pGDE). 2.1 Decision-making strategy In classical decision theory, a decision is the selection of an action from a set of alternative actions. Decision theory does not have much to say about actions–neither about their nature nor about how a set of alternative actions becomes available to the decision maker. A decision is good if the decision maker believes that the selected action will prove at least as good as the other alternative actions. A good decision is formally characterized as the action that maximizes expected utility, a notion which involves both belief and desirability (Dastani, Hulstijn, & Van der Torre, 2005). If the decision making is based on the human reasoning pattern, it is known as practical reasoning. Practical reasoning is the decisionmaking or thinking of rational entities (Rens, 2010).

The second type of information is used by AGVG in the autonomy control module ACM, see Figure 1. The autonomy control module is constituted by a classical control loop. An autonomy control loop is a process management system designed to maintain an FMS performance variable at a desired result (set-point). That is, the control loop focuses on improving a process variable (FMS performance) through vDE autonomy control. Since the autonomy control loop decides whether to make an adjustment, the adjustment change affects the overall performance. The autonomy control module (ACM) takes as the base the error, that is, the difference between the actual FMS performance and the desired set-point. Then, the ACM output specified by the control algorithm defines the vDE autonomy value according to the FMS performance. The main task of the AGVG is to assign AGVLs to requests for material movements that come from the processing machines. On the other hand, the AGVG passively observe and interpret the actions of AGVL. This method would allow AGVG not only to interpret AGVL team current actions, but also to calculate that AGVL's future actions. In a sense, this method utilizes implicit communication, since the observing AGVL receives information from the actions of the observed AGVL.

Practical reasoning can be divided into deciding what to do and determining how to do it. It explain the actions with terms such as know, think, believe, want, need, prefer, goal, desire, should, able to, impossible, intend, plan and act or action. It infer the mental state or mental attitude by observing the behavior and labeling the attitude with such a term. And then, it communicate the mental states using these terms(Rens, 2010). In the context of practical reasoning, there are many conceptualizations and formalizations of decision making. However, two models that have been widely used in the implementation of autonomous entities are, the beliefdesire-intention (BDI) model and the markov decision processes (MDPs) model. Markov decision process is based

Besides, the AGVG can provide globally decisions because the GDL has a global view of the AGV system. The AGVG 9

2019 IFAC IMS 10 August 12-14, 2019. Oshawa, Canada Sergio Ramiro Gonzalez et al. / IFAC PapersOnLine 52-10 (2019) 7–12

on Markov‘s Property. The BDI model has come to be possibly the best known and best studied model of practical reasoning entities. Because, BDI models focus less on the definition of the optimal decision represented by the decision rule, but instead also discuss the way decisions are reached(Dastani et al., 2005). Also, BDI entities are able to balance the time spent on deliberating about plans (choosing what to do) and executing those plans (doing it) (Dastani et al., 2005) (Georgeff, Pell, Pollack, Tambe, & Wooldridge, 1998).

leaves via the loading/unloading node located at node (M1). Three assembly workstations(𝑀𝑀2 , 𝑀𝑀3 , 𝑀𝑀4 ) equipped with Kuka robots and an automated inspection unit 𝑀𝑀5 are placed around a transportation system. In the experimental case, the FMS can manufacture several products by mounting raw components to form letters (e.g., B, E, L, T, A, I and P).

BDI models implements the principal aspects of Michael Bratman's theory of human practical reasoning whose main elements are the belief, desire, intention and plan (as mental states), to explain some aspects of human behavior. Beliefs (B), desires (D), and intentions (I) are the three primary mental attitudes and they capture the informational, motivational, and decision components of an entity, respectively(Rens, 2010). In this tradition, beliefs represent the information of an entity has about the state of the world. Belief is like knowledge, except that it does not have to be true. Desires or goals represent the preferred states of affairs for an entity, in an ideal world, would wish to be brought about (Dastani et al., 2005). Intentions are the chosen means to achieve the entity's desires, and are generally implemented as plans and post-conditions. BDI model, are applied in for example natural language is to formalize the internal architecture of complex entities. It is a way of explaining future-directed intention, and has been applied as a way of limiting the time spent deliberating on what to do by eliminating choices inconsistent with current intentions.

Figure 2 Original FMS layout; blue circles represent each original node of AIP Primeca

3.2 The test method In this section, a testing method is described to simulate the vDE autonomy behaviour given the states (input variables) and to check the correct FMS performance. To validate the proposed approach, an instantiation is implemented using multi-agent system (MAS) technology in the Netlogo® program. Traditionally, FMS problems have been solved by analytical methods, but the simulation method has gained popularity, as it enables the modelling and analysis of complex FMS. Simulation presents an excellent tool for visualizing, understanding and analysing the dynamics of manufacturing systems and thus assisting in the decisionmaking process.

BDI model has been applied in this approach for the decision-making of intelligent AGV entities. An AGV entitiy is able to continuously reason about beliefs, goals, and intentions and act accordingly(Shajari & Ghorbani, 2004). The pGDE creates a list of desires to pursue and focus decides on a subset of desires to seriously pursue as goals. When the intention is available from AGV’s, it uses to get a plan. Periodically dropping invalid intentions and instantiating new intentions as necessary in the current dynamic situation, with the situation’s particular rate of change. The application to BDI reasoning allows reducing the action decision time by eliminating inconsistent choices relative to the intention. 3

The implemented system was tested using an emulated version of the real AIP-PRIMECA FMS, and the scenarios contained in this work were defined taking into consideration the ones specified in the benchmark in (Trentesaux et al., 2013). To illustrate the idea behind our approach, a layout of a vDE transport system is proposed and represented, as shown in Figure 3. This model includes the design and operational issues served by AGVLs moving within a looplike network environment while supporting bidirectional material flows. The topology of the vDE transportation system is assumed to be associated with a strongly connected graph. This graph contains nodes, which are disjunction points (red circles), and the black dashed points (links). The links are the parts of the system that require no decisions during the transport process (guide path), since the AGVL can only move in one direction towards the next node. The green arrows indicate AGV bidirectional routes. In these experiments, all the machines are available for performing each processing part from the real FMS; besides, the FMS is equipped with three identical AGVLs for material handling purposes. The AGVs would receive the raw material from the loading unit and start visiting the different workstations to complete the production cycle. In the experimental case, the FMS can manufacture only products by mounting raw components to form letters B, E, L, T, 𝐴𝐴, 𝐼𝐼 𝑎𝑎𝑎𝑎𝑎𝑎 𝑃𝑃

EXPERIMENTAL CASE STUDY

This section relates the simulation studies carried out in the AIP-PRIMECA assembly cell to evaluate our approach. In this section a robust testing methodology to ensure that the approach being developed has been fully tested to confirm that it meets the specified requirements and can successfully operate in the FMS environment. 3.1 The FMS model The original manufacturing system AIP-PRIMECA FMS of Valenciennes (France) has seven workstations, placed around a conveyor belt system with transfer gates, which employs self-propelled shuttles to transport the products along the track’s blue line (see Figure 2). Each product enters and 10

2019 IFAC IMS August 12-14, 2019. Oshawa, Canada Sergio Ramiro Gonzalez et al. / IFAC PapersOnLine 52-10 (2019) 7–12

11

(MTTR). The time between failures on a particular machine or AGV is generated from an exponential distributed mean time between failures (MTBF) of 20s (Roser, Nakano, & Tanaka, 2002). This is a modest model but it usually offers a good description of the incidence of exceptional events. It is possible to include more disturbances, but, for our case, only those already mentioned before are taken into account. 4

RESULTS

All the experiments and simulations were performed on a desktop PC with Intel Core-i5 2.4GHz, 8G Ram and Windows 7 OS. The simulation process starts by selecting an XML file that is generated in Jupyter from Python software that describes the simulation to be processed. This file describes the number of times and the scenario to be processed. For each scenario described, one result file (Excel file) is created, containing the achieved results, which can be processed and analysed later. The simulation was run for approximate 12 minutes each scenario. The experiments for the FMS model described in 3.3 are concentrated in the design of the AGV system to enable them to execute their tasks in a flexible and robust manner and in the interaction between the AGVL and the resources from the FMS. However, the evaluation of the semi-heterarchical control with respect to the system performance is rather difficult. This is due to the fact that we need a decision-making execution strategies such as BDI model for the autonomy control when it want to execute the task to be solved.

Figure 3 AGV graph proposed.

In this approach, all the AGVLs are equipped with a decision-making process with the capability to execute their task autonomously when needed. In performing their assigned tasks, the AGVs need to make a decision, which calls the related method for assigning the best AGV for dispatching products (task). In this approach, dispatching is a procedure for assigning AGVs to products. A dispatching system is a scheduling system with a zero-planning horizon, and a dispatching decision is made when an AGVL drops off a load, a vehicle reaches its parking location and a new load arrives. In our test, three sets of simulations were carried out, see next section. 3.3 The disturbances proposed FMS environments are dynamic in nature and subject to various unpredictable disturbances or breakdowns, because they are designed to work over a long period of time without any interruption. Hence, these conditions either are not valid or will potentially take the system out of the established overall performance. Disturbances of machines and equipment in FMSs can lead to heavy losses. In our approach, our goal is to measure and control the AGVL autonomy as criteria for adapting to the disturbances intelligently and effectively. This ability should keep the manufacturing system running and avoid manufacturing process delays and decay of the system performance. We constructed some scenarios with disturbance specification features. These scenarios are used: no breakdown, vDE breakdown and machine breakdown. They contain features that also occur in real-world situations. We have one standard scenario, no breakdown, to test whether the AGV and machine resources perform equally well without disturbances. vDE reflects an AGV disturbance situation; in this scenario, the failure is distributed equally over the AGV fleet. Machine breakdown is proposed with FMS machine disturbances; in this scenario, the failure is over only one processing machine. It assumes that machines and AGVs can only break down while they aren’t operating. The adaptation to these disturbances involves the ability to respond rapidly and to recover autonomously. The machines or AGVs stops for a lapse of time, simulating a technical failure according to scenarios. The elapsed and repair times are defined by the exponential distribution. When a breaks down, it must be repaired, and it takes either 24, 25, or 26 seconds for the repair to be completed, according to the exponential distributed mean time to repair

800 Performance values

758,1 596,9

600

Machine Breakdown

650 449,69 343,04 269

400 159,46 78,71 79,01

200

AGV breakdown 127,05 40,93 40,93

0 Makespan

Autonomy

MSD

CTV

Figure 4. The comparison result of scenarios with 2 vehicles

The experimental results see Table 1 and Figure 4, illustrated that the real time obtained by global control increased the autonomy value as criteria for adapting to the disturbances intelligently and effectively. The semi-heterarchical control of AGV adjustable autonomy has a significant impact on the performance of the system (material processing measures): due date mean square deviation (MSD), completion time variance (CTV) and Makespan (Cmax). Each point on the Table 1 represents the experimental data on the relevant level. It is implied that the autonomy results in improvements in FMS performance. Then, the heterarchical architecture is a factor is liable to be declared to have a significant effect on autonomy and FMS performance measures. This is due to the fact that the AGV interaction effect in this case is highly correlated with the AGV autonomy. The vDEs’ autonomy is based on indirect characteristics, such as negotiations and communications based on (Alonso, Fuertes, Martinez, & Soza, 2009). The 𝑤𝑤𝑐𝑐= 0.360 and 11

2019 IFAC IMS 12 August 12-14, 2019. Oshawa, Canada Sergio Ramiro Gonzalez et al. / IFAC PapersOnLine 52-10 (2019) 7–12

𝑤𝑤𝑛𝑛= 0.12 are the weight values of each characteristic of communications and negotiations, respectively. N is the number of intentions, and C is the number of communications. The autonomy value is calculated with the algebraic summation of the aforementioned weight values multiplied by the number of intentions and communications. Autonomy level = C ∗ wc + N ∗ wc .

6

This work would not have been possible without the Pontificia Universidad Javeriana PUJ and CEIBA Fundation for their guidance and the financial support they provided me during these years. Certainly, this support has allowed me to focus on the development of my research. Besides, I would like to acknowledge the collaborative work with the “Centro de automatizacion industrial” (CTAI) laboratory located at PUJ Bogota-Colombia. I extend my deepest gratitude to each of them.

Table 1 Performance measures of the control proposed Case of study No disturbances AGV breakdown Machine breakdown

Number of AGVs 1 2 3 1 2 3 1 2 3

Makespan (seconds) 1654.8 650.0 795.5 1771.3 596.9 619.8 1720.5 758.1 1659.4

5

Autonomy

MSD

CTV

10.11 79.01 72.68 10.45 159.46 154.86 10.11 78.71 125.13

1005.70 343.04 537.90 917.40 269.00 359.94 1080.27 449.69 1386.14

2893.19 40.93 22.85 2273.73 127.05 21.08 2754.80 40.93 33.10

ACKNOWLEDGMENTS

7

REFERENCES

Alonso, F., Fuertes, J. L., Martinez, L., & Soza, H. (2009). Towards a set of Measures for Evaluating Software Agent Autonomy. 2009 Eighth Mexican International Conference on Artificial Intelligence, 73–78. https://doi.org/10.1109/MICAI.2009.15 Baker, A. D. (1998). A survey of factory control algorithms that can be implemented in a multi-agent heterarchy: dispatching, scheduling, and pull. Journal of Manufacturing Systems, 17(4), 297–320. Barber, K., & Martin, C. (1999). Agent autonomy: Specification, measurement, and dynamic adjustment. Proceedings of the Autonomy Control Software Workshop at Autonomous Agents, 1999, 8–15. Barber, K. S., & Martin, C. E. (2000). Autonomy as decision-making control. International Workshop on Agent Theories, Architectures, and Languages, 343–345. Springer. Castelfranchi, C. (1994). Guarantees for autonomy in cognitive agent architecture. International Workshop on Agent Theories, Architectures, and Languages, 56–70. Springer. d’Inverno, M., Luck, M., & others. (1996). Understanding autonomous interaction. ECAI, 529–533. Wiley. Dastani, M., Dignum, F., & Meyer, J.-J. (2003). Autonomy and agent deliberation. International Workshop on Computational Autonomy, 114–127. Springer. Dastani, M., Hulstijn, J., & Van der Torre, L. (2005). How to decide what to do? European Journal of Operational Research, 160(3), 762–784. Dorais, G., Bonasso, R. P., Kortenkamp, D., Pell, B., & Schreckenghost, D. (1999). Adjustable autonomy for humancentered autonomous systems. Working Notes of the Sixteenth International Joint Conference on Artificial Intelligence Workshop on Adjustable Autonomy Systems, 16–35. Falcone, R., & Castelfranchi, C. (2001). The human in the loop of a delegated agent: The theory of adjustable social autonomy. IEEE Transactions on Systems, Man, and Cybernetics-Part A: Systems and Humans, 31(5), 406–418. Farahvash, P., & Boucher, T. O. (2004). A multi-agent architecture for control of {AGV} systems. Robotics and ComputerIntegrated Manufacturing, 20(6), 473–483. http://dx.doi.org/10.1016/j.rcim.2004.07.005 Gabriel Zambrano Rey. (2014). Reducing Myopic Behavior in FMS Control: A Semi-Heterarchical SimulationOptimization Approach. Universitate de Valenciennes et du Hainaut- Cambresis,. Georgeff, M., Pell, B., Pollack, M., Tambe, M., & Wooldridge, M. (1998). The belief-desire-intention model of agency. International Workshop on Agent Theories, Architectures, and Languages, 1–10. Springer. González, S. R., Mondragón, I., Zambrano, G., Hernandez, W., & Montaña, H. (2017). Manufacturing Control Architecture for FMS with AGV: A State-of-the-Art. In I. Chang, J. Baca, H. A. Moreno, I. G. Carrera, & M. N. Cardona (Eds.), Advances in Automation and Robotics Research in Latin America: Proceedings of the 1st Latin American Congress on Automation and Robotics, Panama City, Panama 2017 (pp. 157–172). Retrieved from https://doi.org/10.1007/978-3-319-54377-2_14 Gunderson, J. P., & Martin, W. N. (1999). Efects of uncertainty on variable autonomy in maintenance robots. Workshop on Autonomy Control Software. Autonomous Agents, 99. Hentout, A., Messous, M.-A., & Bouzouia, B. (2015). Fault-tolerant multi-agent control architecture for autonomous mobile manipulators: Simulation results. Computers & Electrical Engineering, 43, 238–256. Hülsmann, M., & Windt, K. (2007). Understanding autonomous cooperation and control in logistics: the impact of autonomy on management, information, communication and material flow. Springer Science & Business Media. Jensen, K., & Kristensen, L. M. (2009). Coloured Petri nets: modelling and validation of concurrent systems. Springer Science & Business Media. Jimenez, J.-F., Bekrar, A., Trentesaux, D., & Leitão, P. (2016). A switching mechanism framework for optimal coupling of predictive scheduling and reactive control in manufacturing hybrid control architectures. International Journal of Production Research, 54(23), 7027–7042. https://doi.org/10.1080/00207543.2016.1177237 Kumar, K. D., Karunamoorthy, L., Roth, H., & Mirnalinee, T. T. (2005). Computers in manufacturing: towards successful implementation of integrated automation system. Technovation, 25(5), 477–488. http://dx.doi.org/10.1016/j.technovation.2003.09.004 KUSTAK, A. (1985). Flexible manufacturing systems: a structural approach. International Journal of Production Research, 23(6), 1057–1073. Leith, D. J., Shorten, R. N., Leithead, W., Mason, O., & Curran, P. (2003). Issues in the design of switched linear control systems: A benchmark study. International Journal of Adaptive Control and Signal Processing, 17(2), 103–118. Maheswaran, R. T., Tambe, M., Varakantham, P., & Myers, K. (2003). Adjustable autonomy challenges in personal assistant agents: A position paper. International Workshop on Computational Autonomy, 187–194. Springer. Meystel, A., & others. (2000). Measuring performance of systems with autonomy: metrics for intelligence of constructed systems. White Paper for the Workshop on Performance Metrics for Intelligent System, NIST, Maryland, USA, 14–16. Nandula, M., & Dutta, S. P. (2000). Performance evaluation of an auction-based manufacturing system using coloured Petri nets. International Journal of Production Research, 38(10), 2155 – 2171. Park, H.-S., & Tran, N.-H. (2012). An autonomous manufacturing system based on swarm of cognitive agents. Journal of Manufacturing Systems, 31(3), 337 – 348. https://doi.org/10.1016/j.jmsy.2012.05.002 Pynadath, D., Scerri, P., & Tambe, M. (2001). MDPs for adjustable autonomy in real-world multi-agent environments. 2001 AAAI Spring Symposium, Technical Report SS-10-03, Stanford, CA, AAAI. Rao, A. S., Georgeff, M. P., & others. (1995). BDI agents: from theory to practice. ICMAS, 95, 312–319. Rens, G. B. (2010). A belief-desire-intention architechture with a logic-based planner for agents in stochastic domains (PhD Thesis). Roser, C., Nakano, M., & Tanaka, M. (2002). Utilization vs. Throughput: Bottleneck Detection in AGV Systems. Logistic System Symposium. Scerri, P., Pynadath, D. V., & Tambe, M. (2002). Towards adjustable autonomy for the real world. Journal of Artificial Intelligence Research, 17, 171–228. Schurr, N., Marecki, J., Lewis, J. P., Tambe, M., & Scerri, P. (2005). The defacto system: Training tool for incident commanders. AAAI, 1555–1562. Shajari, M., & Ghorbani, A. A. (2004). Application of Belief-Desire-Intention Agents in Intrusion Detection & Response. PST, 181–191. Towards adjustable autonomy for the real world. (2002). JOURNAL OF ARTIFICIAL IN℡LIGENCE RESEARCH, 17, 171 – 228. Trentesaux, D., Pach, C., Bekrar, A., Sallez, Y., Berger, T., Bonte, T., … Barbosa, J. (2013). Benchmarking flexible jobshop scheduling and control systems. Control Engineering Practice, 21(9), 1204–1225. van der Vecht, Bob, Dignum, F., Meyer, J.-J. Ch., & Neef, M. (2008). A Dynamic Coordination Mechanism Using Adjustable Autonomy. In J. S. Sichman, J. Padget, S. Ossowski, & P. Noriega (Eds.), Coordination, Organizations, Institutions, and Norms in Agent Systems III: COIN 2007 International Workshops COIN@AAMAS 2007, Honolulu, HI, USA, May 14, 2007 COIN@MALLOW 2007, Durham, UK, September 3-4, 2007 Revised Selected Papers (pp. 83–96). Retrieved from https://doi.org/10.1007/978-3540-79003-7_7 Vecht, B. (2009). Adjustable autonomy: Controling influences on decision making. Utrecht University. Verhagen, H. J. (2000). Norm autonomous agents (PhD Thesis). Citeseer. Vuković, N., & Miljković, Z. (2009). New hybrid control architecture for intelligent mobile robot navigation in a manufacturing environment. FME Transactions, 37(1), 9–18.

CONCLUSIONS

In this paper, an architecture decentralized with a BDI model is used by AGVGs to make their control decisions. When entities have decision-making control over increasing periods of time, it makes them less dependent and thus more autonomous (Verhagen, 2000). One of the key elements found in this paper in the future is the possibility of influencing and transferring control capabilities between entities, allowing rapid control responses under limited information based on temporal assessment of the environment. A set of experiments was illustrated to demonstrate the applicability and the effectiveness of the semi-heterarchical architecture proposed. The semi-heterarchical control module was implemented in this paper. In these circumstances, the next challenge is to develop an autonomy control module for switching between local (heterarchical) and global (semi-heterarchical) approaches according to the autonomy requirements and FMS performance. This module should include and combine the reactivity to environmental changes, scalability, robustness against the occurrence of disturbances, easier integration of manufacturing resources and intelligence capabilities. In the future work, the plan is to analyse the transport throughput time to improve the efficiency of our solution. We also plan to validate it on harder problems, even the ones that require investigation with a large feature offset. In the future work, the plan is to analyse the transport throughput time to improve the efficiency of our solution. We also plan to validate it on harder problems, even the ones that require investigation with a large feature offset.

The main contributions of this paper are to investigate in more detail the AGV transport observed following the application of autonomy control. This is performed through semi-heterarchical architecture to control of AGV autonomy that can provide the flexibility to operate and make decisions, also ensuring the reactivity and adaptability needed to deal with perturbations. The use of control in the management of autonomy emphasizes the need for examination of the response to mitigate the perturbations’ impact. 12