A Hybrid Evolutionary and Multiagent Reinforcement Learning Approach to Accelerate the Computation of Traffic Assignment

Traditionally, traffic assignment allocates trips to links in a traffic network. Nowadays it is also useful to recommend routes. Here, it is interesting to recommend routes that are as close as possible to the system optimum, while also considering the user equilibrium. To compute an approximation of such an assignment, we use a hybrid approach in which an optimization process based on an evolutionary algorithm is combined with multiagent reinforcement learning. This has two advantages: first, the convergence is accelerated; second, the multiagent reinforcement learning resembles the adaptive route choice that drivers perform in order to seek the user equilibrium. In short, our hybrid approach aims at incorporating both the system and the user perspectives in the traffic assignment problem. Results confirm that this hybridization accelerates the computation and delivers an efficient assignment.