Affiliations: Computer Science Department, LIIAN Laboratory, Faculty of Science Dhar-Mahraz, Sidi Mohamed Ben Abdellah University, Fez, Morocco
Correspondence:
[*]
Corresponding author: Chaimae El Hatri, Computer Science Department, LIIAN Laboratory, Faculty of Science Dhar-Mahraz, Sidi Mohamed Ben Abdellah University, Fez, 30000, Morocco. E-mail:chaimae.elhatri@usmba.ac.ma
Abstract: In this paper, a novel traffic management model
is presented, which simultaneously optimizes vehicle re-routing and traffic
light control to alleviate traffic congestion and limit the effects of
incidents on traffic flow based on Multi-Objective Particle Swarm
Optimization (MOPSO) method. Once a congested road is predicted, our
proposed Multi-Objective Traffic Light Control is then applied to optimize
signal timing which takes the maximization of traffic flow on the edge where
the incident takes place and the minimization of the average junction
waiting time as two objectives. To improve the performance and sensitivity
of MOPSO algorithm, we used Q-Learning algorithm to grant to each agent of
the swarm the ability of selecting appropriate MOPSO parameters adapted to
the structure of the problem. At the same time, when the situation of the
traffic flow starts to become more serious, we adopt a novel Multi-Objective
Vehicle Re-routing strategy for assigning alternatives routes to cars before
entering the congested road, in order to perform dynamic load balancing.
Vehicle re-routing is also optimized by MOPSO to simultaneously find the
shortest and least popular path. The obtained results from the simulation
using SUMO, a well-known microscopic traffic simulator, confirm the
efficiency of the proposed system.
Keywords: Multi-Objective Particle Swarm Optimization (MOPSO), multi-agent system (MAS), q-learning, traffic light control, vehicle re-routing, Simulator of Urban Mobility (SUMO), JAVA Agent Development Framework (JADE), Traffic Control Interface (TraCI)
