A Novel Self-Learning Intelligent Traffic Signal Control System for Congested Urban Areas

Population is steadily increasing worldwide resulting in intractable traffic congestion in urban dense areas. Adaptive Traffic Signal Control (ATSC) has shown strong potential to effectively alleviate urban traffic congestion by adjusting the signal timing plans in real-time in response to traffic fluctuations to achieve desirable objectives (e.g., minimize delay).The problem of coordinated ATSC is challenging due to the exponential growth in the number of joint timing plans to be explored as the network size grows. A novel Multi-Agent Reinforcement Learning for Integrated Network of Adaptive Traffic Signal Controllers (MARLIN-ATSC) system is designed, developed and evaluated at a large-scale urban network of 59 intersections in the lower downtown core of the City of Toronto for the morning rush hour. MARLIN-ATSC offers two possible modes: (1) independent mode, i.e. each intersection controller is independently working of other agents; and (2) integrated mode, where each controller coordinates the signal control actions with the neighbouring intersections. The MARLIN-ATSC large-scale application was conducted to examine two cases: 1) corridor-specific agent coordination, and 2) network-wide agent coordination. In the network-wide coordination experiments three scenarios were investigated: uniform demand profile, variable demand profile, and unfamiliar drivers (using a low percentage of familiar drivers). The results show unprecedented reduction in the average intersection delay ranging from 27% in mode 1 to 39% in mode 2 at the network level; and travel time savings of 15% in mode 1 and 26% in mode 2, along the busiest routes in downtown Toronto.