Learning in Cooperating Agents Environment as a Method of Solving Transport Problems and Limiting the Effects of Crisis Situations

Abstract

Abstract. The realising of transport requests plays an important role for companies regarding cost. For this reason the construction of an optimal and effective transport planning and scheduling, offering the best use of transport means is hugely important. One of the researched transport problems at present is PDPTW. In this work, the PDPTW problem will be extended by adding changeable and uncertain travel times between given locations, which will then be examined and learnt by the dispatching company system. The changeable travel times are the results of traffic jams forming and propagating. The applied multi-agent approach allows us to consider the problem on the level of a whole company and on the local level analysed by a particular vehicle. The other aspects which are taken into consideration are traffic patterns, whose goal it is to facilitate the optimal routes for the given traffic patterns. These patterns contain the information about the dependencies between the traffic state on the particular routes.

Multi-agent Crisis Management in Transport Domain

“…In [11] a general idea of the model of such a system to react on crisis situations, prevent them and minimise consequences is given. The crisis situations analysed are most often delays caused by uncertain travel times and traffic jams [7]. Furthermore, the presented MARS system was equipped with a module responsible for modelling and predicting traffic jams while constructing routes.…”