Multi-Agent Systems as a Platform for VANETs
TIER 3 TIER 2
norm-governed system, where a norm is a rule which prescribes or-ganizational concepts like (institutional) power permission, obligation, sanction, and other more complex relations. This approach to multi-agent systems has its roots in the study of legal, social and organizational systems, while formal anal-yses often use some version of deontic logic which provides route to automation. Such automation offers formal defini-tions of responsibility and control, interoperability (as de-fined in terms of external, executable specifications), and reasoning about sub-ideal states of the system (i.e. detect-ing and recoverdetect-ing from faults in the system, which are to be expected in probabilistic systems like VANETs.) Therefore, our solution is to propose a two-layer agent archi-tecture comprising: 1)‘lightweight’, network-facing, mobile agents; 2) ‘heavyweight’, application-facing, norm-aware agents;
and 3) a common language between the two types of agent.
The mobile agent layer supports in implementation deci-sions made by norm-aware agents. Thus we propose to con-verge ‘heavyweight’ agents, which operate in the domain of norms and ‘codes of conduct’ about the network, with lightweight mobile agents which operate in the domain of network-centric events and parameters in provisioning the network itself. In the following sub-section, we elaborate further on this novel network architecture.
3.2 Network Architecture
Considering all the properties of the agent systems, we pro-pose to implement the two-layer agent architecture over a three-tier network architecture as shown in figure 1. The first tier contains the vehicles which are communicating them-selves on the road which are within a given cluster (a VANET).
A cluster is formed based on radio range coverage of vehicles and the road-side base station. In the second tier, a set of clusters are there in which each cluster comprises of base stations (C1 to CN clusters). The clusters may communi-cate by using base stations. The third tier is the transport agency (TA), owned by a private body or government agency to monitor the entire transportation infrastructure and offer the relevant services to the vehicles on the road.
TIER 3
• Vehicle to Base Station: A fixed infrastructure com-prised of (at least) a number of base stations strate-gically positioned in close proximity to the highways is necessary to facilitate the upload/download of data from/to the vehicles. Each base station covers a clus-ter. We assume that several sensors information are input to the base station. Information could be traf-fic density, vehicle types, adverse road conditions, etc.
The agency in the base station comprises of following components: Base Station Manager Agent (BSMA), Service Agent (SA), Advertisement Agent (ADA) and Cluster Knowledge Base (CKB).
– BSMA: It is a static norm-governed agent de-ployed at each base station which maintains and synchronizes all the agents that are associated with base station. It regularly updates the CKB with the visited vehicles and its services infor-mation by interacting with VMA of each vehicle.
Also computes the traffic density maps, adverse road conditions and updates the CKB. Critical information received from VMAs in its cluster is sent to other BSMAs. This agent is responsible for communicating information with VMAs, BS-MAs and TA.
– SA: It is a static agent responsible for collecting the services information from the service providers of the cluster and regularly updates the CKB. It also broadcasts any critical information available with it to the vehicles within its cluster upon no-tification from other BSMAs.
– ADA: It is a mobile agent which roams in the network and informs the visited VMAs about the auctions, special exchange offers, ticket reserva-tions, etc. It may interact with the user and get the information about his participation in auc-tions or booking tickets or any such tasks as the user wishes.
– CKB: It comprises of information such as critical events within cluster, services available in cluster, visited vehicle information, traffic density maps, road conditions, location aware services, adver-tisements, etc.
• Cluster to Cluster: Clustering provides a method to disseminate the traffic information as well as provide varieties of services. Whole network is divided into self-managed groups of vehicles called clusters. These clusters continually adapt themselves to the chang-ing network topology and new cluster configurations.
Communication between the clusters will take place with the help of BSMAs located in base stations that are fixed on the road side, although, as discussed below it is possible to manage clusterign without BMSAs.
• Cluster to TA: TA consists of complete information of the transportation infrastructure which is accumulated from various cluster BSMAs. BSMAs of the clusters communicate with the TA manager. TA manager pe-riodically constructs a overall picture of the road ways in terms of traffic, critical events, road conditions, etc., and constructs a road map and distributes to the BS-MAs. It also prepares list of services available in its
entire area and stores in its knowledge base which may be used by SDA to discover the services.
Practical implementation of the proposed scheme needs the following: 1) the vehicle must be equipped with a computa-tional device comprising a real time operating system, wire-less transceiver unit with dynamic ranges, GPS unit, speed sensing unit, inter-vehicle distance monitoring unit, cam-eras (optional), fuel sensing unit, human interface, embed-ded tyre air sensing unit, database manager, an agent plat-form with set of static and mobile agents; 2) the base station must have a computational unit, wireless transceiver unit, real time operating system, agent platform, cameras and database manager; 3) environment and road condition sen-sors are connected to base station; and 4) Transport agency comprises of computational unit, wireless transceiver unit with dynamic ranges, real time operating system, database manager, agent platform, human interface and Internet con-nection.
A further refinement of the network architecture, facilitated by the use of agent technology, is this. Instead of fixed base stations situated at strategic points along the highway, each defining a cluster, and vehicles belong to a cluster according to proximity to a base station; weremovethe base stations altogether (with a few exceptions), and thelogical clusters nowphysicallymove the length of the highway, and moving vehicles join or leave clusters according to their ground speed and proximity to identified cluster-heads or gateway nodes.
The additional research questions that need to be addressed now include:
• Permanent transience (or transient permanence): how the network stays ‘the same’, even though every net-work node is different (by analogy, someone is the
‘same person’, from one year to the next, even though every cell is different);
• Role-based and policy based network management: who (vehicle node VMAs) gets which role, (e.g. as cluster-head, or gateway, etc.) and why, on what basis, and so on. In other words, some VMAs have to assume the responsibility and functionality of BSMAs. For this we need elections etc. (cf. [23]);
• Anticipation: knowing when a network change is im-minent due to role hand-off (vehicle leaving the high-way), and to take pre-emptive behavior to ensure the continuous smooth-running of he network. This be-haviour could be based on a cognitive characterization (BDI (Belief Desire Intention)-like) of the mental state of the agent (VMA) [24].
4. APPLICATION SCENARIOS
In this section we illustrate the operation of the system to re-alise three of the application scenarios mentioned earlier. We assume that an agent platform exists in vehicles, base sta-tions and TA. However agents communicate with each other by using traditional exchange mechanisms if an agent plat-form is not available in any of the components of VANET.
The agent platform provides following services: agent cre-ation, mobility, communiccre-ation, security and fault tolerance.
4.1 Access to Fixed Infrastructure
Access to fixed infrastructure is essentially using the VANET to connect to any computer terminal in the car to the Inter-net; however, we seek to optimise performance by caching regularly-accessed information in the cluster. Required in-formation is first searched within the cluster, i.e., by polling the BSMA and the VMAs. If the information is not avail-able within the cluster, it searches in the neighboring clus-ter. The expected information cached would be road maps, traffic density maps, articles, etc.; of course other Internet services e.g. VoIP would also be available.
VMA SDA
VMA
VMA
SDA
Cluster 1 Cluster 2
BSMA BSMA
VMA
Neighbour 1
5
7 SDA
SDA SDA
4 2
3
6
8
9 SDA
Figure 2: Information access in VANETs The proposed information access model is shown in figure 4.1.
The method to access the information is as follows given in sequence. It is assumed that information required is avail-able in the vehicle of neighboring cluster.
1. The vehicle needing information contacts the BSMA through VMA. BSMA searches information in its CKB and also contacts TA. In this case, BSMA as well as TA does not have the information, hence it informs the VMA.
2. VMA creates the SDA to it’s neighboring vehicles.
3. SDA migrates to neighboring vehicle and communi-cates to VMA through TKB.
4. If the required information is available in TKB of the neighboring vehicles SDA sends the information to the VMA.
5. If the information is not available with the neighboring vehicle, SDA clones from its place and moves to second degree neighbors and so on within the cluster. If it identifies the required information with a particular vehicle then the information will be sent to VMA. SDA and its clone destroy themselves once they reach the end of the cluster.
6. In case if information is not available within the clus-ter, VMA again generates SDA which migrates to its base station.
7. SDA clones itself to neighboring clusters by communi-cating with its cluster BSMA under certain norms.
8. In case SDA gets information at neighboring BSMA, it returns to its created VMA.
9. If SDA fails in getting the information from neighbor-ing BSMAs, it searches within the neighborneighbor-ing cluster vehicles as given in steps 3 to 5.
4.2 Critical Information Dissemination
Search for information, in the above application, is con-cerned with information pull. In this application, we are concerned with information push, whereby vehicles spread messages about safety related events such asaccidents, road conditions (roadworks), inter-vehicle distance, weather con-ditions ahead, etc., through AA. Critical information related events may be of two kinds. Firstly, there are events (such as fuel status, vehicle speed, neighbor vehicle distance, etc.) that can be detected by an AA for a particular vehicle.
These events will assist the driver in safer driving and it does not need to be spread to other vehicles. Secondly, there are events such as traffic jams, accidents, road conditions, etc., which have to be disseminated to other vehicles in an ag-gregated way. Aggregation requires aggregating the events sensed by a single vehicle as well as aggregating the events of all the vehicles.
VMA
VMA
VMA
BSMA BSMA
VMA AA
1
AA
4
Cluster 1 Cluster 2
2
3
5 VMA
VMA
Figure 3: Critical information dissemination model The critical information dissemination using proposed model is shown in figure 3. The method to disseminate the critical information follows the given sequence:
1. Whenever critical events occur, VMA in a vehicle cre-ates AA to its neighbors. AA communicates with neighboring VMA and informs about the critical event as well as collects any critical information available in the visited vehicle.
2. The neighboring VMA which received the message of critical event, creates clones of AA based on certain norms and spreads the message to its neighboring ve-hicle and so on. In this way the message is reached
to all the vehicles within the cluster as well as AAs aggregate the critical event information and pass on the information to its VMA. All the cloned agents de-stroy themselves once they move out of the range of its cluster BSMA.
3. VMA communicates critical information to its BSMA.
BSMA updates CKB based on certain permitted ac-tions depending on the norms.
4. BSMA communicates about the received critical infor-mation to the neighboring BSMAs as well as to TA.
5. Neighboring BSMAs broadcasts critical information in its cluster.
4.3 Location-Dependent Services
Location-dependent services can be built over the informa-tion push-pull model of the two previous scenarios. Cer-tain information such as the location of the nearest facilities like fuel stations, parking zones, entertainment places and restaurants, markets, etc., can be accessed through TA. Fig-ure 4 depicts the information ‘advertisement’ by using the proposed model. The method to access local information about roadside (or nearby accessible) services is then given by the following sequence.
BSMA SA
2
3
4
TA manager 1
VMA ADA
ADA
5
VMA
Figure 4: Advertising information from TA 1. TA manager sends the information to all BSMAs.
2. BSMA updates its CKB and informs SA to advertise the message through ADAs.
3. SA creates several ADAs which move to the nearest vehicles and pass on the information to users as well as interact to get some or nil response for the adver-tisement.
4. ADAs clones themselves and visits all the vehicles that are not visited by any other ADA and repeats the op-eration as given in step 3.
5. Parent ADA accumulates all the responses and sends the information to SA which in turn passes on the information to TA manager.
5. BENEFITS OF USING AGENTS
The following are some of the benefits of using agents in the proposed vehicular information ad hoc networks:
• Flexibility: Agents are flexible to accommodate vari-eties of services to facilitate information dissemination in VANETs. For example, SDA may be encoded to dis-cover multiple services rather than single service based on user degree of satisfaction.
• Adaptability: As we observe in the applications men-tioned above, we can see that agents such as AA, SDA and ADA adapt to varied network conditions such as vehicle mobility, occurence of critical events, changes in road and weather conditions, etc.
• Maintainability: The agent-based approach we have advocated is predicated entirely on the idea of open systems, that is, the interaction of heterogeneous and unpredictable components. However, the use of norm-aware agents considers situations where design-time specifications may need to be modified at run-time; or where the system specifications may only be partially given at design-time, and the components themselves complete the specifications at run-time. This is an entirely new approach to network management.
• Survivability: Wireless networks are specifically de-signed to operate in the expectation of contention and error. Similarly, in VANETs, it may be that a node fails to comply with the system specifications, either by design, by accident or from necessity. Dealing with such non-compliant behavior, can also be addressed by the norm-governed approach, where appropriate behavior can be stipulated using concepts stemming from the study of legal and social systems: e.g. permis-sions, obligations and other normative relations such as power, right and entitlement.
6. CONCLUSIONS
Vehicular ad hoc networks provide an exciting area of re-search at the intersection of a number of disciplines and tech-nologies. There is a good future for applications of VANET, ranging fromdiagnostic, safety tools, information services, and traffic monitoring and management to in-car digital en-tertainment and business services. However, for these appli-cations to become everyday reality an array of technological challenges need to be addressed.
In this position paper, we have outlined an agent architec-ture for VANETs inspired by previous work in QoS (Qual-ity of Service) provisioning in MANETS [23], in which we developed a 3-layer system architecture integrating mobile agents and norm-aware agents to deliver a rapid response with rational deliberation about social behaviour, contracts and sub-ideal situations. This solution was based on in-tegrating ‘lightweight’, network-facing, mobile agents with
‘heavyweight’, application-facing, norm-aware agents, via a common language so that the mobile agent layer supports in the network those decisions made by the norm-aware agents.
The paper addressed the use of emerging agent technology in VANETs. It can be assumed that multi-gent systems have a great potential to influence the design of future VANET and
their services. Multi-agent systems should be regarded as an
“add on” to existing service platforms, providing more flexi-bility, adaptaflexi-bility, and personalization for the realization of services within next generation VANET environments. We are planning to implement the proposed work by using IBM aglets workbench as well as simulate using NS2 to evaluate the performance of the system.
Acknowledgements
Jeremy Pitt was supported by Royal Society Science Net-work No. 16751 and UK EPSRC Grants GR/S69252 and GR/T20328. We appreciate the reviewers’ useful comments.
7. REFERENCES
[1] J. Jetcheva, Yih-Chun Hu, S. PalChaudhuri, et al.,
“Design and Evaluation of a Metropolitan Area
Multitier Wireless Ad Hoc Network Architecture”,Proc.
5th IEEE Workshop on Mobile Computing Systems and applications (WMCA), Monterey, CA, Oct. 2003 [2] Jun Luo and J.-P. Hubaux, “A Survey of Inter-Vehicle
Communication”,Proc. Embedded security in Cars-Securing current and Future Automotive IT applications, pp. 164-179, Springer-Verlag, Oct. 2005 [3] A. Ebner, H. Rohling, M. Lott, R. Halfmann,
“Decentralized Slot Synchronization In Highly Dynamic Ad Hoc Networks”,Proc. 57th IEEE Vehicular
Technology Conference, Jeju, South Korea, 2003.
[4] M. Rydstrom, A. Toyserkani, E. Strom, A. Svensson.
“Towards a Wireless Network for Traffic Safety Applications”,Proc. Radio and Communication, pp.
375-380, Linkoping, Sweden, 2005
[5] S. Goel, T. Imielinski, K. Ozbay, “Ascertaining Viability of Wi-Fi based Vehicle-to-Vehicle Network for Traffic Information Dissemination”,Proc. 7th Annual Intelligent Transportation System Conference, 2004.
[6] S. S. Manvi, P. Venkataram, “Applications of agent technology in communications: A review”,Computer communications, vol. 27, pp. 1493-1508, Oct. 2004.
[7] Jiancong Chen, S.H. Gary Chan, Jingyi He, Soung-Chang Liew, “Mixed-Mode WLAN: The Integration of Ad Hoc Mode with Wireless LAN
Infrastructure”,Proc. IEEE International Conference on Communications (ICC), Korea, 16-20 May, 2005
[8] K. Zimmermann, L. Egger, M. Brunner,
“Self-Management of Wireless Base Stations”,Proc.
IEEE Workshop on Management Issues and Challenges in Mobile Computing, Nice, France, 2005.
[9] P. Sethi, G. Barua, “Dynamic Cluster Management In Ad hoc Networks”,Proc. HPC 2002, Bangalore, December 2002.
[10] M. Bechler, H.J. Hof, D. Kraft, F. Pahlke and L. Wolf,
“A Cluster-Based Security Architecture for Ad Hoc Networks”,Proc. IEEE Infocom, Hong Kong, China, March 2004.
[11] F. Sailhan, V. Issarny, “Scalable Service Discovery for MANET”,Proc. 3rd IEEE International Conference on Pervasive Computing and Communications, 2005.
[12] M. Al-kahtani, H. Mouftah, “Congestion control and clustering stability in wireless ad hoc networks:
Enhancements for clustering stability in mobile ad hoc networks”,Proc. 1st ACM Workshop on Quality of service & security in wireless and mobile networks, 2005 [13] M. Jakobsson, Xiao Feng Wang, S. Wetzel, “Stealth
Attacks in Vehicular Technologies”,Proc. First International Workshop on Vehicular Ad-hoc Networks, Philadelphia, USA, Oct. 2004
[14] S. Eichler, F. Dotzer, C. Schwingenschlogl, F. Caro, J. Eberspacher, “Secure Routing in a Vehicular Ad Hoc Network”,Proc. IEEE 60th Vehicular Technology Conference, Los Angeles, USA, 2004.
[15] S. Jaap, M. Bechler, L. Wolf, “Evaluation of Routing Protocols for Vehicular Ad Hoc Networks in City Traffic Scenarios”,Proc. 11th Open European Summer School EUNICE, Colmenarejo, Spain, July 2005.
[16] M. Nekovee,“Sensor networks on the road: the promises and challenges of vehicular ad hoc networks and grids”,Proc. Workshop on Ubiquitous Computing and e-Research, National e-Science Center, Edinburgh, UK, May 2005.
[17] H. Fubler, M. Moreno, M. Transier, A. Festag, H. Hartenstein, “Thoughts on a Protocol Architecture for Vehicular Ad-hoc Networks”,Proc. 2nd International Workshop in Intelligent Transportation (WIT 2005), pp.
41-45, Hamburg, Germany, March 2005.
[18] I. Athanasiadis and P. Mitkas,“An agent-based intelligent environmental monitoring system”,Proc.
Management of Environmental Quality, pp. 238-249, 2004.
[19] G. Cabri, L. Leonardi, M. Mamei, F. Zambonelli,
“Mobile Agent Organizations”,Proc. Fifth IEEE International Conference on Mobile Agents (MA), Atlanta, Georgia, USA, 2001.
[20] T. Magedanz, K. Rothermel, S. Krause, “Intelligent Agents:An Emerging Technology for Next Generation Telecommunications?”,Proc. IEEE Infocom, pp. 24-28, San Francisco, CA, USA, 1996
[21] S. S. Manvi and P. Venkataram, “An agent based adaptive bandwidth allocation scheme for multimedia application”,Journal of systems and software, vol. 75, no. 3, pp. 305-318, 2004
[22] A. Jones, M. Sergot, “On the characterization of law and computer systems: The normative system
perspective”, in J. Meyer and R. Wieringa (eds),Deontic Logic in Computer Science, John Wiley, 1993.
[23] J. Pitt, P. Venkataram, A. Mamdani, “QoS management in MANETs using norm-governed agent societies”,Proc. 6th International Workshop Engineering societies in the agents’ world, Kusadasi, Turkey, 2005.
[24] C. Castelfranchi, E. Lorini. “Cognitive Anatomy and Func-tions of ExpectaFunc-tions. InProceedings of IJCAI03 Workshop on Cognitive Modeling of Agents and Multi-Agent Interac-tions, Acapulco, Mexico, August 9-11, 2003.