Introducing joint trips in a multi-agent transport simulation: from agents to clique replanning

Abstract

MATS im is a multi-agent transport simulation software, based on an iterative relaxation procedure. In this framework, each agent is assigned a daily plan consisting of a sequence of trips and activities, to which a score is associated. Daily plans are optimised knowing the state of the traffic flows. The system iterates between traffic flow simulation and plan modification, until an equilibrium is reached. In this report, the ability for the agent to travel together with other agents is introduced. This is done by passing from optimisation of the scores of individual plans to the optimisation of scores associated to joint plans. To perform such an optimisation, a genetic algorithm is used, which is presented. This algorithm is able to improve the scores of a joint plan by identifying pertinent joint trips. However, clues show remaining convergence difficulties, which are analysed.