Multi-Agent Environment for Modelling and Solving Dynamic Transport Problems

Abstract

The transport requirements in modern society are becoming more and more important. Thus, offered transport services need to be more and more advanced and better designed to meet users demands. Important cost factors of many goods are transport costs. Therefore, a reduction of costs, a better adjustment of strategies to the demand as well as a better planning and scheduling of available resources are important for the transport companies. This paper is aimed at modelling and simulation of transport systems, involving a dynamic Pickup and Delivery problem with Time Windows and capacity constraints (PDPTW). PDPTW is defined by a set of transport requests which should be performed while minimising costs expressed by the number of vehicles, total distance and total travel time. Each request is described by two locations: pickup and delivery, periods of time when the operations of pickup or delivery can be performed and a load to be transported. The nature of this problem, its distribution and the possibility of using a lot of autonomous planning modules, lead us to use a multi-agent approach. Our approach allows the modeling of entities which do not appear in the classical PDPTW such as company organisation, communication among vehicles, interactions between vehicles and company dispatcher or different strategies of requests acceptation by different vehicles. This paper presents also a software environment and experimentations to validate the proposed approach.