AGENT-BASED AND EVOLUTIONARY PLANNING TECHNIQUES SUPPORTING CRISES MANAGEMENT IN TRANSPORTATION SYSTEMS Rafal Dre˙ zewski, Marek Kisiel-Dorohinicki, Jaroslaw Ko´ zlak Institute of Computer Science AGH University of Science and Technology Al. Mickiewicza 30, Kraków, Poland e-mail: {drezew,doroh,kozlak}@agh.edu.pl Abstract: The paper deals with a particular approach to crises management in transportation systems. The considerations are based on a layered reference architecture dedicated to monitoring and management of multi-agent systems. In the contribution agent-based and evolutionary approaches are proposed as a basis for the simulation layer realisation. Several variants are shortly discussed and illustrated by selected experimental results. Copyright c  2006 IFAC Keywords: transportation systems, multi-agent systems, crisis management. 1. INTRODUCTION During the last decade the idea of an intelligent au- tonomous agent gains more and more interest both in academic community and in industry. A constantly increasing number of computer systems are being analysed and designed in terms of agents. Agents play a key role in integration of AI sub-disciplines, which is often necessary to design and build modern intelligent systems. Agent technology is used in various domains, providing concepts and tools for development of com- plex, distributed and decentralised systems. The systems under consideration may both be de- signed from scratch as multi-agent ones (operating in the virtual world, e.g. network information services, virtual enterprises), as well as function in the reality as a set of cooperating autonomous subsystems of whatever origin (e.g. transportation systems, industrial complexes). Acceptance of the agent-based approach opens possibility for solving many problems that until 1 This work was partially sponsored by State Committee for Scien- tific Research (KBN) grant no. 3 T11C 025 27 now has been tractable only with respect to tightly coupled centralized systems. Some of these problems are risk and critical situations (states) analysis (Wu and Soo, 1999; Collins et al., 1999). The problem critical situations analysis with respect to transportation systems will be of our special interest here. Obviously effective construction of a transport planning allows companies to highly limit sustained costs and be more competitive on the market. There- fore, an important challenge is to create tools, which support development of such planning on the basis of acquired knowledge on available transport resources, incoming transport requests and road network struc- ture. One of the applied approaches is simulation re- search, which facilitates selection and configuration of transport planning algorithms. Based on a general scheme of crises management in multi-agent systems (MAS), as well as preliminary results obtained in the field of transportation systems (Nawarecki et al., 2005), a variety of possible vari- ants of planning-support techniques are considered in the paper. Section 2 introduces fundamental concepts