Estimating the performance of intelligent transport systems wireless services for multimodal logistics applications

Expert Systems with Applications 39 (2012) 3939–3949

Contents lists available at SciVerse ScienceDirect

Expert Systems with Applications journal homepage: www.elsevier.com/locate/eswa

Estimating the performance of intelligent transport systems wireless services for multimodal logistics applications Adrian E. Coronado Mondragon a,⇑, Etienne S. Coronado Mondragon b, Christian E. Coronado Mondragon c, Franklin Mung’au d a

School of Management, Royal Holloway University of London, Egham, Surrey, UK Networking and Telecommunications Professional Services, Montréal, QC, Canada School of Ocean Technology, Marine Institute, Memorial University of Newfoundland, St John’s, NL, Canada d Logistics Institute, The University of Hull, Cottingham Road, Hull, UK b c

a r t i c l e

i n f o

Keywords: Network design solution Multimodal logistics Vehicle to infrastructure data transfer performance Service provisioning Dedicated short range communications (DSRC)

a b s t r a c t Various wireless technologies including radio frequency identification (RFID), bluetooth, cellular networks and dedicated short range communications (DSRC) might have an impact on logistics and transport operations. Among them, DSRC stands out as a broadband communications technology which has been designed to provide a general purpose Radio-Frequency (RF) link between vehicles and network infrastructure. As such, DSRC is capable of implementing the physical layer of an Internet Protocol (IP) bearer based network designed to facilitate the monitoring and coordination of portside vehicular traffic. This unprecedented application of wireless networking has the potential to greatly enhance the management of the flow of goods and resources, particularly within large, international ports whose activities comprise multimodal operations such as the use of road haulage to move cargo transported by sea. Given the need for reliable services in non-safety business applications, in this work an Intelligent Transport Systems (ITS) approach is used to address two issues. First, in wireless networks reliable data transfer transport layer services are affected where there is an apparent increase in mobility when access point coverage areas are reduced to counter the effects of path loss in the physical layer. Second, a service provisioning protocol intended for vehicle to infrastructure (V2I) data transfer is proposed to illustrate the importance of cumulative costs in wireless networks used for logistics applications. The analysis covers the average response time for requesting on-demand services within the portside network considered. The results of the analysis confirm the suitability of the approach used to provide a logistics network capable of meeting the requirements demanded in multimodal logistics. Ó 2011 Published by Elsevier Ltd.

1. Introduction Mobile wireless communications have evolved rapidly over the last 40 years; from low capacity systems based entirely on analogue techniques and delivering voice only services, to high capacity, digitized systems providing a variety of services including voice and data. The popularity and growth of mobile wireless communications has been mainly due to the ability of the technology to provide high data rate communications whilst adhering to stringent Quality of Service (QoS) requirements; in part due to advancements in wireless channel modelling techniques and the subsequent development of sophisticated digital transmission methods (Akaiwa, 1997).

⇑ Corresponding author. Tel.: +44 (0) 1784414348; fax: +44 (0) 1784 276100. E-mail addresses: [email protected] (A.E. Coronado Mondragon), [email protected] (E.S. Coronado Mondragon), christian.coronado@mi. mun.ca (C.E. Coronado Mondragon), [email protected] (F. Mung’au). 0957-4174/$ - see front matter Ó 2011 Published by Elsevier Ltd. doi:10.1016/j.eswa.2011.08.161

Research on transportation has investigated information propagation in traffic streams using inter-vehicle communications. For example, Jin and Recker (2006) discussed the reliability of inter-vehicle communication in a traffic stream dependent on the distribution of equipped vehicles. By assuming that information propagation is instantaneous compared to vehicle movements, the authors measured reliability as the probability of success for information to travel beyond a location. Wang’s (2007) research studied information propagation along a traffic stream where equipped vehicles follow an independent homogeneous Poisson process. The results from that work demonstrated the relationship between propagation distance, equipped vehicle density and transmission range. The work by Blythe (2005) recognises that Mobile Adhoc Networks or other wireless devices comprise the intelligent infrastructure required to enable vehicles to be constantly in communications with other vehicles near them as well the infrastructure which can then deliver location based services and intelligent control and safety applications. Recent wireless vehicle networks

3940

A.E. Coronado Mondragon et al. / Expert Systems with Applications 39 (2012) 3939–3949

developments such as DSRC have improved the reliability of not only V2V but also V2I communications. Moreover, in recent times sophisticated modelling tools have facilitated the design and analysis of complex wireless networks. Nonetheless, new challenges have risen in terms of the need to provide expedient secure services of the highest quality with minimum delays over optimized network designs. Given the increasing popularity of network-based services and the need for assure of interoperability, roaming, and end-to-end session management, Huang, Chen, Chen, and Wu (2009) proposed a service-oriented information dissemination scheme for vehicle infotainment system to ensure the possibility of heterogeneous vehicular communication and a framework for delivering real-time services over an IP-based network. Fuzzy logic was used to assist in building a reliable and robust communication network environment that showed acceptable levels of performance including average latency and dropping probability. A recent development in the mobile communications technology market has been a shift to ‘‘all IP’’ networks because equipment and installation costs, as well as operation and maintenance costs are being driven downwards by evolving IP technologies (Janevski, 2003). In the current rate of proliferation of mobile communications and in particular mobile IP networks, it is perhaps inevitable, and indeed expedient, that mobile wireless communications are deployed extensively for logistics (Coronado, Lalwani, Coronado, & Cherkaoui, 2008). It is expected that logistics, like other areas in intelligent transportation, would benefit extensively from adopting current advances in mobile communications. Mobile wireless communications are a fundamental component of ITS. Indeed, it has been acknowledged that ITS are based in the use of advanced ICT to achieve a reduction of congestion and accidents while making transport networks more secure by reducing their impact on the environment (ERTICO Research Project., 2007). In logistics, ITS play an important role in achieving paperless information flows, efficient traffic management by the use of Automatic Identification and Data Capture (AIDC) and tracking and tracing by using satellite positioning services (Zomer & Anten, 2008). Logistics is an important knowledge-based economic activity that plays a major role worldwide. Logistics deals with everything involving planning, organising and managing activities that provides goods or services (Anonymous, 1997). In Europe, logistics expenditure is in the order of €800 billion whilst representing 7% of GDP and with transportation costs representing 43% of the total logistics expenditure (Zomer & Anten, 2008). According to the European Commission the trend in market share of freight transport modes comprises 45% road haulage and 41% sea transportation (Browne, Allen, & Woodburn, 2006). Advances in mobile communications are suitable for complex logistics arrangements that result from the combination of different sorts of transportation like road, sea and air. For example, port operations are a complex arrangement in which road transportation can be seen as a feeder to sea transport. In Europe, sea routes (short sea) between neighbouring countries today offer high quality regular services that can be combined with other transport modes to provide efficient alternatives (Anonymous, 2007). The amendment to the IEEE 802.11 set of standards for wireless local area networks (WLAN IEEE 802.11p) which adds wireless access in a vehicular environment (WAVE) gives the opportunity to develop innovative applications and services that can be used in logistics and transportation. The 802.11p standard is set to pave the way to enhancing existing dedicated short range communications (DSRC) standards (Anonymous, 2008), which provide wireless channel specifications for roadside to vehicle (vehicle to infrastructure, V2I) and vehicle to vehicle (V2V) communications environments. DSRC technology operates in the Super High

Frequency (SHF) Band at 5.9 GHz, where radio waves propagate mainly in the Line-of-Sight as well as due to multipath propagation (Zhao, Kivinen, Vainikainen, & Skog, 2002). By operating at 33 dBm (2W), DSRC is expected to provide coverage over a range of up to 1000 m with a data rate up to 27 Mbps (NHTSA, 2005) per channel (including two control channels and seven service channels). However, this maximum specified communications range may not be realised in all cases of DSRC technology deployment as, for example, the portside environment presents a significant challenge due to obstructions such as warehousing and stacks of shipping containers which can adversely affect radio reception. Indeed, the received signal power can fall of with distance raised to the fifth power (50 dB per decade) due to the resulting multipath propagation and path loss phenomenon with significant repercussions on the network application. DSRC is designed to handle the transmission of both safety and non-safety messages into two modalities: vehicle to vehicle (V2V) and vehicle to infrastructure (V2I). Within this technology two types of messages are transmitted: Wireless Access for Vehicular Environment (WAVE) Short Messages (WSM) and IPv6 traffic (WAVE, 2005). WSM messages involves low latency and critical safety-related messages assuming a realtime propagation while IPv6 traffic is generally related to commercial services such as download or streaming of data. In the particular context of logistics applications, which are motivated by the need for efficiency in logistics and visibility and transparency in the supply chain, we propose the adoption of a ubiquitous wireless networks in which vehicles transmit periodic information updates that can be interpreted by higher level applications. Data transfers that meet the needs of logistics are at the watershed between real-time and elastic network applications, where both reliable data transfer and application response times are required for the efficacy of the application. To counter the effects of multipath propagation and path loss, coverage areas can be reduced with the disadvantage being an increased number of handoffs and increased application latency (Dellapos, De Marco, & Trecordi, 1998) due to Mobile IP (MIP) control message overheads and handoff latency due to the cell switching algorithm (Campbell et al., 2000). The IEEE 1609 (.3) standard describes the transport and network layer services including addressing and routing in support of reliable WAVE data transfer. The IEEE 1609.3 standard specifies the Transmission Control Protocol (TCP) in the transport layer which relies on sequence numbers and acknowledgement to provide a best effort service for end system data transfer. MIP is an internet architecture and protocol for supporting mobility by allowing the mobile user to maintain a single address when moving from one network access point coverage area to another. In so doing, Mobile IP makes the user mobility transparent to the network application, and the user appears stationary for the purposes of data transfer. The current standard specifies the use of indirect routing to the mobile node, which is facilitated by agent discovery and handoff. However, this scheme is characterized by inefficient triangular routing and home agent overloading and route optimization in mobile IP has been proposed as an alternative. The performance evaluation in terms of end-to-end delay for both schemes is provided in the literature (Dellapos et al., 1998), but these results are yet to be extended to the analysis of a ubiquitous, mobile IP network where the offered data traffic pattern is determined by a logistics application. In the particular case of logistics applications, the success in the deployment of service provisioning models for highly mobile environments depends on the implementation of robust architectures capable of maintaining their overall performance when facing hostile environments (Coronado & Cherkaoui, 2007). Perhaps the greatest challenge for this type of service deployment (services offered on the roadside infrastructure likely to be

3941

A.E. Coronado Mondragon et al. / Expert Systems with Applications 39 (2012) 3939–3949

considered on-demand services) is that the exchange of information between users and service providers must be kept reliable and secure, when sensitive information is transferred such as financial transactions or disclosure of user identities. In this case it becomes essential to define a communication model that addresses on-demand service requests enabling the of exchange messages. This paper investigates the response time when executing the tier process for requesting on-demand services within a network using a service provisioning protocol. A service might be represented by a logistics application where the application consists of handshaking and security protocols, as well as the data payload transfer. The paper is relevant at a time transport and logistics companies need to define their policies regarding the adoption of technologies to support their businesses. The layout of the paper is motivated by the observation that the application proposed is unprecedented, and therefore the analysis and simulation of the network performance are described in detail to qualify the conclusions presented. Section 2 describes the importance of WAVE to logistics which then leads to a description of the details of the portside network to support logistics operations in Section 3. In Section 4, a network topology is presented and dimensioned after which the requirements network performance analysis using cumulative signalling costs for on-demand service is provided. In Section 5, the portside network is evaluated using OPNETÒ Wireless Network Modeller (Anonymous, 2009) based on the cumulative costs considerations presented in Section 4. Section 6 provides concluding remarks and it is noted that the simulations results support the notion of the feasibility of DSRC communications and networking for logistics operations.

2. Importance and complexity of multimodal logistics operations A review of Europe’s transport strategy has led to a policy change towards co-modality or multi-modal logistics which is about the efficient use of different modes on their own and in combination, resulting in a sustainable and optimal utilisation of resources (Zomer & Anten, 2008). Complex logistics operations have become multimodal in essence, with a high role played by sea transportation. In the particular case of the UK, almost 95% of imports arrived by sea in 2008 (Anonymous, 2008a) and with road haulage moving a significant amount of cargo in and out of British ports. International logistics requires information and communication (ICT) systems that satisfy a diversity of needs (Leviäkangas, Haajanen, & Alaruikka, 2007). International logistics is practically always multimodal and involves a number of different players who underline the challenge of implementing information services that work to serve the needs of the whole logistics chain (Leviäkangas et al., 2007). Recognising the importance of having a communication platform capable of supporting multimodal transport, Dullaert et al. (2009) developed a solution consisting of a real-time decision support system in which intelligent software agents are used to handle communicative tasks, exchange desired amounts of information among different users using common exchange protocols which act as translators between different systems. We believe that the adoption of mobile communications in IP based networks can have a major impact on the efficiency of multimodal logistics operations especially at a time where government agencies are busy launching initiatives that will contribute towards efficient transportation of goods through a better use of resources. For example, according to the European Commission, in 2005, short sea shipping in the North Sea Region totalled some 591 million tonnes, with regular liner services and ferries

operating fast, reliable and flexible connections that carry a wide range of cargos in a wide range of vessels, including charter vessels that transport bulk steel and construction materials between terminals in the region as well as Roll On–Roll Off (RO–RO) operations. However the same commission has identified road congestion in Europe as a major issue affecting the efficiency of sea transport. In the view of the Commission, the established benefits of short sea shipping as a sustainable part of the logistics chain creates demand for the extension of the North Sea short sea network. The ‘‘Motorways of the Sea programme’’ funded by the European Union (Anonymous, 2007a) is looking at the increase of short sea shipping in the North Sea particularly on infrastructure (port infrastructures, infrastructures for direct land and sea access as well as inland waterway and canal infrastructures) and facilities (electronic logistics management systems, facilities to ensure and enhance safety and security, facilities to simplify administrative and customs procedures). DSRC offers to become a reliable wireless network platform to support advanced logistics applications. In recent years logistics applications supporting location and content information in the form of telematics services have received significant attention in the logistics and transportation sectors. At present, the world market for fleet management, vehicle telematics and communications solutions for professional and private markets is already worth more than €3.9 billion (InnovITS., 2008) and around the world every day, more than 20,000 vehicles are fitted with telematics systems (Practel Inc., 2008). The adoption of logistics and transportation services running on a DSRC platform has great potential given the inherent characteristics of the road haulage industry in Europe. For example, European trucking is predominantly shorthaul (Practel Inc., 2008) and over 85% of truck fleets operate locally with most hauliers moving cargo within their own national borders (InnovITS, 2008). A particular type of service sought is freight efficiency which accounts for 18% of the market for telematics applications in Europe (Practel Inc., 2008). Ports represent locations where the availability of logistics services running on wireless networks such as DSRC technology can have a significant impact on the performance of logistics operations. An example where logistics applications are needed in a portside network would be in situations where the port terminal operator interrogates the trucks (mobile nodes) for track and trace purposes or for updating the records of an application running in the central server based on the status of a job carried out by a truck. Logistics services based on wireless networks such as DSRC technology can offer superior reliability and security compared to current solutions offered in the market based on cellular and satellite systems. Table 1 shows a comparison of the performance for DSRC, cellular and satellite for some of the most important applications currently found in the market such as mileage user fees, probe data, signage, tolls, traffic data and vehicle to vehicle safety (Marousek, Andrews, & Dorfman, 2008). The complexity of logistics operations like those found in ports reflects the importance of having an infrastructure capable of supporting communication accessibility and reliability. Network mobility is a key requirement for achieving ubiquitous connectivity in IP networks deployed in industrial environments like ports. Table 1 Application comparison for wireless technologies (source Marousek et al., 2008). Application

DSRC

Cellular

Satellite

Mileage based user fees Probe date Signage Tolls Traffic data Vehicle to vehicle safety

Good Fair Good Good Fair Good

Fair Fair Poor Poor/fair Fair Unusable

Unusable Unusable Poor/fair Unusable Good Unusable

3942

A.E. Coronado Mondragon et al. / Expert Systems with Applications 39 (2012) 3939–3949

The effect that costs can have in the operation of a portside network to support logistics applications represents a key issue addressed in this work, as these are at the top of the list of issues affecting wireless networks. However, cost analysis can be complex because there are fixed costs in the form of equipment and installation costs, operation costs and maintenance costs. As this work considers a service provisioning protocol intended for V2I data transfer, a structure in the form of cumulative costs is proposed for a comprehensive network performance analysis. This structure of cumulative costs including signalling costs, processing costs and link costs (wired and wireless costs) is used in conjunction with the average response time to confirm the suitability of the logistics network and the service provisioning protocol to support data exchange requirements demanded in modern day logistics/transport operations. 3. Details of the network architecture for portside logistics operations This section describes the functionality of the proposed network to support logistics applications with both the wired and wireless components as well as the reliable data transfer of the data objects used by the logistics network application for a port facility characterized by the use of haulage vehicles and sea transportation. 3.1. The portside network The adoption of DSRC-based network as a platform to support applications for logistics and transport must represent a value proposition on its own. An important activity where the adoption

of a wireless network like DSRC represents a significant value proposition is the movement of bulk material in a port terminal. This study identified a port terminal in the north of England, a port location characterized for the absence of application services that can provide full track and trace capabilities for road haulage vehicles serving the port premises. The use of a logistics application based on the use of wireless vehicle networks can have a significant impact on the performance of logistics operations such as: – Real time track and trace of vehicles and their loads in order to respond to customer enquiries. – Register of accurate haulage traffic entering and moving within the port premises. – Elimination of costly mistakes associated to unloading bulk material in wrong sites which can result in delays to vessel departures (e.g. delays to vessels, depending on size, can cost from $45,000 USD to $150,000 USD a day). – Accurate billing to customers. Some customers demand very close monitoring of products such as bio-mass and fertilizers. – Real-time updates and accuracy on drivers and operators payments (drivers paid by tonnage moved and UK/European road legislation do not apply in the port). – Comply with UK and European road legislation enforced outside port premises. The proposed portside network consists of a local area network with Road Side Unit (RSU) access points connected to core network routers in a star configuration (see Fig. 1). The core routers are connected in a fully meshed or semi meshed network configuration with a maximum number of core connections K, (where K is an integer) such that the number of hops between the routers is

Fig. 1. The portside network with the star configured local access network, the meshed core network, the mobile nodes (haulage vehicles), and the remote work station.

A.E. Coronado Mondragon et al. / Expert Systems with Applications 39 (2012) 3939–3949

minimized. Mobile switching centres depict the ability of the core routers to manage data transfers from several access points and the channelling of this information to the network routers. The network architecture supports three main applications; reliable data transfer between haulage vehicles (deploying onboard units – OBU) and the centralized data repository (reliable data transfer between a remote workstation and the centralized data repository) and reliable data from the remote workstation to the centralized repository. In Fig. 1 an IP cloud represents the data transfers to or from the centralized workstation. Based on the observations made at the port terminal located in the north of England, a practical example of where such network could be deployed is in the allocation of trucks to tipping operations involving bulk materials from a vessel docked in a given berth (point of discharge) to a deposit/warehouse within the confinements of the port perimeter. The same technology can also be applied to the other multimodal operations such as handling of containers and Ro–Ro. The above mentioned data transfer sessions compliment higher level applications that can use the real-time capabilities of the network to monitor and coordinate portside activities. The main logistics activity considered within the port is the transportation of goods, either bulk or containerized, from a ship to suitable storage areas. These items might then be dispatched from the port premises at a later date upon request by the client, but we restrict our analysis to ship discharge activities that takes place within the port, consigning a study of the operation that go beyond the port boundaries to future studies. The number of haulage vehicles required to discharging a ship N(veh) is given by the following expression:

Nðv ehÞ ¼

RðcraneÞ Lðv ehÞ  RTT ðv ehÞ

ð1Þ

where R(crane) [Tonnes per hour] is the rate at which cranes discharge the ship, L(veh) [Tonnes] is the weight that each vehicle can carry, and RTT(veh) [Trips per hour] is the number of trips per hour the haulage vehicle can make; from loading up at the cranes, the time for weighing and to make the trip to the storage area, and finally the time to make the trip back to the crane. For example, if we have an scenario where parameters are R(crane) = 1800, L(veh) = 30 and RTT(veh) = 3 which, from expression 1, gives the number of haulage vehicles assigned to discharging a ship as 10. 3.2. Use of a logistics network application The logistics application supported by the service provisioning protocol consists of a single session in which a TCP connection is requested and established, user id and password information is transmitted and authenticated, and finally the data payload containing the contents and location of the vehicle are transmitted and acknowledged at the receiver. Details of the payload data exchanged can be contained in a basic haulage schema that shows date, contract number, destination, time, tonnage moved, operator identifier, road hauler identifier and current position. Under this scheme, the basic haulage schema will be stored in the vehicle onboard unit (OBU) and the contents exchanged between the vehicle and the roadside units (RSU) distributed within the perimeter of the port. 4. Details of theory for the proposed portside network architecture This section describes the elements of the architecture considered for the DSRC logistics network. Issues considered include application response time when the mobile nodes, represented in

3943

this case by haulage vehicles, update information at a centralized repository (server). The paper proposes a network topology and dimension the network in terms of link data rates. The aim here is to support the expected data traffic associated to port terminal operations. Then later in this section the network performance analysis is presented using cumulative signalling costs based on the use of a protocol for V2I as the portside network is intended to provide spontaneous on-demand service provisioning. 4.1. Elements and considerations of the portside network The idea of generating a proposed network to support multimodal logistics requires the placement of nodes and link connectivity with the aim of interconnecting source and destination nodes. In the specific case of the logistics network, the source nodes are the network access points (RSUs) and the destination node is the server site. Core nodes (routers) are placed for the purpose of inter-connectivity, that is, to allow for communications between the different local access areas. The decision on where to place nodes and links interconnecting the located nodes is made on an economical basis (i.e., taking into account the fixed cost associated with installing network nodes and the linear, distance-based costs associated with link connectivity). The general premise of the logistics network is ubiquitous network access, which implies that the track and trace operation is possible through periodic information updates of the haulage vehicle at any given location within the network coverage area. It is also assumed that the core of the network is fully-interconnected (fully-meshed) or partially meshed such that hop count is minimized. The node location and link connectivity design problem is adopted from Pioro & Medhi, 2004, pp. 213, 221 and shown in appendix A, this well acknowledged binary linear programming problem is solved using the tree search algorithm due to the small scale of the project (Boyce, Farhi, & Weischedel, 1973). The length and width considered for deployment of the network is 1400 m long by 800 m wide. This area includes the terminals where ships dock in the berth to get discharged, the warehouse/depot terminals where trucks move the material from the ships and the route between the loading and unloading terminals. In the network envisaged for the port facilities, access sites represent RSU locations which are located at regular intervals to provide ubiquitous network access to roaming DSRC equipped mobile nodes. It is desirable that the placement of router/switch locations must consider the physical obstructions present in a site. Once the network topology has been realized, the next step is to determine the required link data rates, a process referred to as network dimensioning (Pioro & Medhi, 2004, p. 263). The solution to the link capacities problem is to determine the link weights that can be used by the well known routing protocols such as the Open Shortest Path First (OSPF) protocol and the Intermediate System to Intermediate System (IS–IS) protocol, that result in the minimum utilization and thus lead to minimal nodal delays (see Appendix B). Unlike the node location and link connectivity network design problem, the Bounded Link Delay network design problem associated to the proposed network is not a linear programming problem because of the implicit relation comprising the vector of network flows. However, meta-heuristic algorithms such as simulated annealing local search can be employed to solve the problem (Pioro et al., 2002). Fig. 2 shows the proposed network as it will look deployed within the confinements of the port terminal area using demand flows and the considerations described above. In Fig. 2 the link loads are the upstream data traffic flows which are derived from the estimation that 48 kbps of data flow from each core site towards the centralized server site. This data traffic pattern estimate is derived by considering that a queue of up to 10 vehicles might develop in all network access site simultaneously,

3944

A.E. Coronado Mondragon et al. / Expert Systems with Applications 39 (2012) 3939–3949

Fig. 2. The proposed portside network over a considered geographical area.

and that each vehicle is communicating data objects (45 kB), with information updates received from all vehicles every 5 min, and that each core router is connected to a maximum of 4 access sites. Clearly, this is a pessimistic estimate for the queuing delays for messages at the routers and actual figure in the practical network should be less than those derived in this analysis. Assuming that the offered data traffic is at the estimated levels, and T1 core network links and 10 Base-T Ethernet local access network links are deployed, the resulting queuing delay at the bottleneck server router is 0.1 ls and a queuing delay of 0.7 ls at the network core routers for each bit transmitted. Fig. 2 depicts the network design stages for an access point separation of 340 m, a suitable separation access point considered in this paper given that the nominal communications range for DSRC equipment is 300 m (Anonymous, 2009a), and this study evaluates a coverage area 40 m over that nominal range.

4.2. Requirements of network performance analysis and cumulative signalling costs The performance analysis of the portside network is based on the utilisation of cumulative costs. Rather than providing a network analysis based on derived analytical expressions of well known methods for estimating reliable data transfer, this subsection introduces network performance analysis using cumulative signalling costs based on the use of a V2I protocol. In the particular case of the portside network supporting a logistics application it is important to consider that robustness and scalability determine the deployment of reliable information services. The portside network, geographically restricted to the port facility studied, also represents the domain architecture with the task of validating and granting network and service access to spontaneous ‘on-demand’ requesters. A session is created when a

A.E. Coronado Mondragon et al. / Expert Systems with Applications 39 (2012) 3939–3949

3945

Fig. 3. Portside network elements affected by the proposed cumulative cost structure.

request is made to access the logistics application – service-run in the portside network. For the purpose of analyzing cumulative costs in the portside network, there are few other modules that need to be considered, apart from the OBU and RSU already mentioned in previous sections of this paper. One module involved is Security (SEC) which is used to identify vehicles as well as for generating secure cryptographic material used in the identification of individual service sessions (Coronado & Cherkaoui, 2007). Another module used is the Session Manager (SM) which is responsible for creating a session ID and for collecting all the session parameters before being forwarded to the RSU. A session ID is required as the mobile nodes (haulage vehicles) need to access services as they perform a job. The service provider (SP) is presented as the server hosting the logistics application used in the portside network. In order to analyze the communication model related to the portside network, it is possible to generalize the cumulative signalling cost based on the sum of individual cost at each tier process. In the particular case of the portside network, the elements considered cover the vehicle onboard unit (OBU), roadside unit (RSU), session manager (SM), security module (SEC) and service provider (SP). Each element is subject to a cost structure of cumulative costs that covers signalling costs, processing costs and link costs (wired and wireless costs). Fig. 3 depicts the elements of the portside network affected by the cumulative costs structure. Table 2 enlists the related processing costs for performing data processing, cryptographic operations and the related link costs

Table 2 Cryptographic, link and processing costs. Symbol

Definition

wl wd

Wireless link cost Wired link cost

C1 C2 C3 C4 C5

Data access processing cost OBU data access processing cost SP data access processing cost RSU data access processing cost SM data access processing cost SEC data access processing cost

bq bx ba bw

Cryptographic cost OBU’s cryptographic cost RSU’s cryptographic cost SP’s cryptographic cost SEC’s cryptographic cost

associated to the elements of the portside network. For simplicity purposes it is assumed the same link cost for all the network elements in the fixed network. Data access processing costs make reference to the costs associated to processing a service request by a mobile node, in this case a haulage truck. To better understand the implications of data processing, Fig. 4 illustrates the exchange of messages necessary to request access to services between the potential user (a mobile node, truck) and the service provider. From Fig. 4, a transitory OBU (user) like those found in the haulage trucks sends an initial message to the RSU notifying its presence as denoted in arrow 1. Then, the RSU replies to the message by requesting the OBU’s attributes and expecting to receive the OBU’s details (e.g. key certificate) shown in arrows 2 and 3. The RSU contacts the Service Provider (SP) based on the service identifier in order to retrieve the SP’s attributes which include the SP’s key certificate through arrows 4 and 5. The collected information from both the OBU and SP is sent to the session manager where a session ID is generated (arrow 6). Once this is done, the SEC module receives the request in arrow 7 where key certificates need to be verified if those do not possess a revoked status. As soon as this validation is retrieved, this module is able to generate the session key for the OBU and the provider, as well as, the corresponding pseudonyms. At this point, all the parameters generated for the session are collected, protected and sent back to the corresponding recipients which are both the OBU and the SP, respectively (arrows 8–11). The use of key certificates becomes important as within the port facilities there can be haulage vehicles belonging to different haulage companies, doing different jobs but allocated to the same contract operation of unloading a vessel. Based on the flow diagram in Fig. 4, the signalling cost at each tier process considers the number of times received messages are processed; additionally, it also considers if cryptographic operations are involved at that tier and the related cost due to wireless or wireline links. The individual signalling cost at each tier process can be expressed from expression 2 to 7.

C OBU ¼ C 1 þ 2 wl þ bq 



ð2Þ

C RSU ¼ C 3 þ 3 wd þ 2 wl þ bx

ð3Þ

C SP ¼ C 2 þ wd þ ba

ð4Þ



ð5Þ



ð6Þ

C SM ¼ C 4 þ 2 wd C SEC ¼ C 5 þ 2 wd þ bw

3946

A.E. Coronado Mondragon et al. / Expert Systems with Applications 39 (2012) 3939–3949

model at the operational frequency specified as 5.9 GHz (IEEE 802.11p – D5.0. standard) and the data rate is set to 11 Mbps, half the maximum data rate of DSRC.

5.2. Results of the communication model using the cumulative costs proposal

Fig. 4. On-demand service request exchange messages (V2I) – portside network.

Therefore, the general expression to define the total signalling cost for requesting access within a single service district domain can be shown in expression 7.

C Total ¼ C OBU þ C SP þ C RSU þ C SM þ C SEC

ð7Þ

5. Simulation results and discussion This section provides a description of the network simulation settings and results obtained. 5.1. DSRC specification During simulation, specific considerations were followed on the specification of the DSRC-based wireless vehicle network. It is well known that the effects of phenomena (e.g. path loss, shadowing and multipath) on wireless technologies such as DSRC results in variations in received signal strength that adversely affect the throughput of the transport layers protocols (Celandroni & Potort, 2002). Assumptions made in this work include the effects of shadowing being mitigated using a suitable transmission power control scheme and multipath is cancelled using OFDM modulation with an appropriately long cyclic prefix (IEEE 802.11p – D5.0. standard). Path Loss considerations were based on the free space propagation

The on-demand service request exchange message V2I protocol logistics application was assembled using the OPNETÒ Application Characterization Environment (ACE)Ò whiteboard tool. The ACEÒ whiteboard is a robust tool suitable to evaluate the behavior of different tier processes within a simulation networking environment. The assembled model lasts 0.385 s and it is useful to illustrate the potential of the wireless configuration for handling increased volumes of data traffic with minimum degradation levels -critical to maintain high levels of traceability and tracking-. The analysis has considered the cumulative processing time at each stage of the communication model involving the execution of data processing and cryptographic operations. The task response time mentioned in this work is the round trip time it takes for a message originated in the vehicle destined to the application server. The cryptographic processing time values are based on the benchmark speeds given in (Crypto++, 2007) for different cryptographic schemes. For example, the operation schemes for RSA 1024 comprise 1.42 ms for signature; 0.07 ms for verification; 0.07 ms for encryption and 1.52 ms for decryption. These time values are used as references in order to estimate the total response time during a simulated on-demand requesting process. The deployment of the ACEÒ tier process in the topology network will result in the estimation of the total average response time while considering the processing time at each tier process. Digital signatures are based on the RSA cryptographic algorithm with a fixed length of 128 bytes, as well as, the corresponding signature verification processes performed at the tier processes. For the OBU, SP, and SM tiers, it is needed to perform RSA (128 bytes) encryption and decryption operations. Moreover, Hashed Message Authentication Code (HMAC) operations are necessary when generating pseudonyms (32 bits) and for key generation, the DiffieHellman (DH) algorithm can be employed at the SEC tier. It is assumed a maximum packet length of 1024 bytes and TCP as a transport layer. The considered processing time for DH 1024 key generation is 0.44 ms per operation and for HMAC (SHA-1) the processing time is 6.279 ms per byte.

Fig. 5. Average response time for 1, 5, 10 and 20 vehicles – portside network.

A.E. Coronado Mondragon et al. / Expert Systems with Applications 39 (2012) 3939–3949

3947

Furthermore, assumptions made include 20 ms of data processing at each tier stage. That is, every time a message arrives to a specific tier it takes 20 ms to process it. The related propagation time imposed by the wireless and wireline mediums is estimated by simulator statistics results. The type of links deployed in the fixed network is set to T-base 10 for the Ethernet based subnets and DS3 links for the WAN sections. The scope of the simulation is to determine the response time when executing the tier process for requesting on-demand services within a network. There is no intention in this paper to deem the time required for link discovery at lower layers but to estimate the response time when accessing services. Four scenarios were simulated for the proposed portside network. A typical work shift is 8 h with trucks following a trajectory from the bulk depot to the tipping berth and then back to the bulk depot where the material is unloaded. Given the length of 0.385 s of the exchange messages protocol, the first scenario identified the average task response time associated to one truck operating a shift of 8 h. The results show an average task response time of 0.440 s, a delay of 55 ms. Further trials with five trucks gave an average response time of 0.455 s, representing a delay of 70 ms. A scenario with ten trucks allocated to the same contract operation shows an average response time of 0.480 s, a delay of 95 ms. Finally, a scenario with twenty trucks was modelled with results showing an average response time of 0.490 s, a delay of 105 ms and capable of supporting the requirements of non-commercial applications in a port. Fig. 5 shows the results of the simulations for one, five, ten and twenty trucks. The results of the trials show that the proposed architecture for the portside network is capable of achieving negligible levels of delays which by no means compromise the need for continuous track and trace monitoring of haulage operations taking place within the confinements of the port terminal. From the results obtained during the simulations, it was observed that the response time values for process request-response between a vehicle and the service domain are acceptable even when increasing the number of vehicles. During the trials the speed of the trucks was fixed to 30 mph. Further tests for a scenario of 10 trucks operating during 36 h shows an average task response time of 0.480 s, same result as if it operated for 8 h only.

average response time for on-demand requests during a typical work shift of 8 h is acceptable even when increasing the number of vehicles. Moreover, the experienced delays in the average task response time between having one truck and twenty was only 50 ms. These results are important as they can be used as the basis for the development of commercial roadside services which can be accessible, reliable and secure. The components of the on-demand services protocol used in this research have specific functions. For example, the purpose of the security module is to verify the certainty of the holder’s key certificates by using public and private certificate revocation lists. Additionally, the security module is responsible for generating the corresponding security attributes for both the requesting user, in this case the truck, and the solicited provider which in this paper is the port terminal. One of the main parts for service provisioning model deals with the implementation of session managers which are responsible for facilitating the transference of existing and valid session parameters to other session managers. Finally, on the question of the feasibility of DSRC for logistics, the capabilities of these networks have been demonstrated during the simulation whilst taking into account the prevailing wireless channel condition and despite a high degree of mobility of the haulage vehicles. Road haulage still represents the main mode of moving goods in many countries and the availability of services over wireless vehicle networks have the potential to make supply chains more efficient by enabling better track and trace of cargo, better fleet control, better route planning, close monitoring of emissions and truck usage among others. For example, the use of wireless vehicle networks will enable live updates to job and transport schedules based on the status of haulage vehicles or the secure exchange of data from digital tacographs.

6. Conclusions

Node location and link connectivity design problem adopted from Pioro and Medhi (2004, pp. 213, 221)

This paper presents an analysis of reliable data transfer for IP based, DSRC communications network implemented for logistics purposes. The proposed network was designed and dimensioned to reduce the implementation costs and reduce the queuing delays when OSPF or IS-IS shortest path routing are implemented. The coverage area of 340 m used in the trials was perfectly adequate to run the simulation trials using OPNETÒ, as not only it was possible to fully cover the area of the port related to the tipping of bulk material but also to ensure a minimum of queuing delays. It was not only the feasibility of a DSRC communication network to provide full coverage for a geographical area but also the case presented in this paper shows that the deployment of DSRC communications as a portside network addresses a major concern regarding the need to provide robust and secure access to services where reliable delivery of information between vehicles and providers must be guaranteed. The on-demand service request exchange message protocol comprising security entities and a session manager evaluated the average response time for a user, in this case a truck, to request a secure session. The results from the simulations show that the

Acknowledgements A.E. Coronado Mondragon was supported by the UK’s Engineering and Physical Research Council (EPSRC) under grant EP/ F067119/1. Appendix A

Indices i = 1,2, . . ., N j = 1,2, . . ., M Constants P nij

cij gj Kj Variables uij = 1 vij = 1 rj = 1

indices for the access sites to be connected indices for the core sites to be connected number of network routers (P < N) linear distance-based cost of connecting access site i to access site j linear distance-based cost of connecting core site i to core site j cost of installing node j maximum number of access point nodes connected to core node j if access site i is connected to core site j; 0 otherwise. if core site i is connected to core site j; 0 otherwise if site j is chosen as the location of a router; 0 otherwise

3948

A.E. Coronado Mondragon et al. / Expert Systems with Applications 39 (2012) 3939–3949

Objectives Constraints

Minimize Z ¼

N X N X i¼1

nij uij þ

j¼1

N X

r ij gi

ðA:1Þ

j¼1

uij ¼ 1;

d

i ¼ 1; 2; . . . ; N

ðA:2Þ

j¼1 N X

uij  K j r j 6 0;

j ¼ 1; 2; . . . ; M

ðA:3Þ

i¼1 M X

v ij  ðP  1Þrj 6 0;

i ¼ 1; 2; . . . ; N

xdp ðwÞ ¼ hd ;

d ¼ 1; 2; . . . ; D

ðB:2Þ

p

XX

Constraints N X

X

dedp xdp ðwÞ 6 ce ce ;

e ¼ 1; 2; . . . ; E

ðB:3Þ

p

w¼W Unlike the node location and link connectivity network design problem, the Bounded Link Delay network design problem is not a linear programming problem because of the implicit relation x = x(w), where x is the vector of network flows.

ðA:4Þ

j¼1

References

j–i

Akaiwa, Y. (1997). Introduction to digital mobile communication. New York: John Wiley & Sons. Anonymous. (1997). MDC, LogLink/Logistics World, 1997. Anonymous. (2007). Finished vehicle logistics magazine, September 2007. Anonymous. (2007a). Notification of a joint call for tender for motorways of the sea projects in the North Sea Region. http://www.dft.gov.uk/162259/165226/ jointcallnorthseamos2007fin1.pdf. Anonymous. (2008). 5.9 GHz dedicated short range communications (DSRC). http:// www.grouper.ieee.org/groups/ssc32/dsrc. Anonymous. (2008a). Raconteur media. Supply chain management supplement on the times, 5 November 2008, p. 11. Anonymous. (2009). OPNET simulation tool. http://www.mil3.com/home.html. Anonymous. (2009a). DSRC. http://www.kpmg.ca/en/industries/cib/transportation/ documents/PlanningForContingencies.pdf. Blythe, P. T. (2005). Congestion charging: Technical options for the delivery of future UK policy. Transportation Research Part A: Policy and Practice, 39(7–9), 571–587. Boyce, D., Farhi, A., & Weischedel, R. (1973). Optimal network problem: a branch and bound algorithm. Environment and Planning, 519–533. Browne, M., Allen, J., & Woodburn, A. (2006). Developments in Western European logistics strategies in Global Logistics. In Donald Waters (Ed.), New directions in supply chain management. London: Kogan Page. Campbell, A., Gomez, J., Kim, S., Valko, A., Wan, Chieh-Yih, & Turanyi, Z. (2000). Design implementation and evaluation of cellular IP. IEEE Personal Communications, 42–49. Celandroni, N., & Potort, F. (2002). Maximizing single connection TCP goodput by trading bandwidth for BER. International Journal of Communications Systems, 16, 63–79. Coronado, E., Cherkaoui, S. (2007). Secure service provisioning for vehicular networks. In International workshop for ITS Ubiroads, co-located with IEEE GIIS 2007, July 2–6 Morocco. Coronado, A., Lalwani, C., Coronado, E., Cherkaoui, S. (2008). Wireless vehicular networks to support road haulage and port operations in a multimodal logistics environment. In Proceedings of IEEE international conference on service operations and logistics, and informatics (pp. 550–555), Beijing, 12–15 October. Crypto++, Crypto++ 5.5 Benchmarks. (2007). http://www.cryptopp.com/ benchmarks.html. Dellapos, M., De Marco, M., Trecordi, V. (1998). Performance evaluation of mobile IP protocols in a wireless environment, conference records. In IEEE international conference on communications (Vol. 3, pp. 507–514). June 1998. Dullaert, W., Neutens, T., Vanden Berghe, G., Vermeule, T., Vernimmen, B., & Witlox, F. (2009). MamMoeT: An intelligent agent-based communication support platform for multimodal transport. Expert Systems with Applications, 36(7), 10280–10287. ERTICO Research Project. (2007). http://www.ertico.com/en/about_ertico/ mission__vision/mission__vision.htm. Huang, C. J., Chen, Y. J., Chen, I. F., & Wu, T. H. (2009). An intelligent infotainment dissemination scheme for heterogeneous vehicular networks. Expert Systems with Applications, 36(10), 12472–12479. InnovITS. (2008). Market analysis of the intelligent transport systems and services (ITSS), UK. Janevski, T. (2003). Traffic analysis and design of wireless IP networks. Artech House. Jin, W.-L., & Recker, W. W. (2006). Instantaneous information propagation in a traffic stream through inter-vehicle communication. Transportation Research Part B: Methodological, 40(3), 230–250. Leviäkangas, P., Haajanen, J., & Alaruikka, A. M. (2007). Information service architecture for international multimodal logistic corridor. IEEE Transactions on Intelligent Transportation Systems, 8(4), 565–574.

N X

uij þ

i¼1

M X

v ij ¼ N þ PðP  1Þ

ðA:5Þ

j¼1

The node location and link connectivity network design problem presented above is a binary linear programming problem that is solved using the tree search algorithm.

Appendix B In this case, the network design objective function is related to the delay per packets at a node, and it is assumed that arrivals are Poisson distributed at a rate ye [pkts/s] and link transmission rates are to be fixed at ce [pkts/s]. An additional consideration is that for OSPF and IS–IS, constraints are imposed on the weight system w and the set of all allowable weights is denoted by W; usually a set of monotonically increasing numbers from a minimum to maximum. The network design problem that solves for the node delays is summarized below (Pioro & Medhi 2004, p. 263): Indices d = 1,2, . . ., D p = 1,2, . . ., PD e = 1,2, . . ., E Constants hd dedp = 1 ce

ce Variables we xdp(w) = 1

data traffic demands network paths for demand d links demands volume if link e belongs to path p realizing demand d; 0 otherwise capacity of link e utilization factor for link e metric of link e, w = (w1,w2, . . ., wE) flow induced by link metric system w for demand d on path p

Objectives

Minimize

E 1X ye E e¼1 ce  ye

A.E. Coronado Mondragon et al. / Expert Systems with Applications 39 (2012) 3939–3949 Marousek, J., Andrews, S., Dorfman, M. (2008). A comparison of communications systems for VII. In Proceedings of the 15th world congress on intelligent transportation systems, November 16–20, New York, USA. NHTSA. (2005). DOT HS 809 859 nhtsa. Vehicle safety communications project task 3 final report. Identify intelligent vehicle safety applications enabled by DSRC, March. Pioro, M., & Medhi, D. (2004). IP routing flows and capacity design in communications and computer networks. Morgan Kaufmann. Pioro, M., Szentesi, J., Harmatos, A., Juttner, A., Gajowniczek, P., & Kozdorowski, S. (2002). On open shortest path first related network optimisation problems. Journal of Performance Evaluation, 201–223. Practel. (2008). Wireless communications for ITS: Technologies and markets, issue II. USA: Practel, Inc.

3949

Wang, X. (2007). Modeling the process of information relay through intervehicle communication. Transportation Research Part B: Methodological, 41(6), 684–700. WAVE. (2005). Wireless access in vehicular environments (WAVE) networking services. Draft IEEE P1609.3/D13, April. Zhao, X., Kivinen, J., Vainikainen, P., & Skog, K. (2002). Propagation characteristics for wideband outdoor mobile communications at 5.3 GHz. IEEE Journal on Selected Areas in Communications, 20, 507–514. Zomer, G., Anten, N. (2008). Vision on the crucial role of ITS for efficient and green European logistics. In Proceedings of the 15th world congress on intelligent transportation systems, November 16–20, New York, USA.