A self-learning traffic system means less waiting at intersections.
Despite the smart technology that has made today’s transportation more efficient, our current traffic lights still operate on an outdated system of timed sequences and ineffective sensors that do not respond to the real-time ebb and flow of traffic. The University of Toronto’s Dr. Samah el-Tantawy has developed MARLIN-ATSC (Multi-agent Reinforcement Learning for Integrated Network of Adaptive Traffic Signal Controllers), an artificial intelligence traffic system that uses cameras and inter-system collaboration to determine how best to manage traffic flow throughout a region. MARLIN’s success is largely due to its adaptive nature, which allows it to learn and adjust as it gathers traffic pattern information.
MARLIN has proven successful in virtual tests, cutting travel times by 25 percent and reducing delays at intersections by 40 percent. el-Tantawy is currently working with Professor Baher Abdulhai, director of the Toronto Intelligent Transportation Systems Centre, to put MARLIN through field tests and on the road to commercialization.