A Traffic Signal Co-learning Adaptive Control Method Based on Gridding Model and Probability Grey Number Theory

The increasing volume of traffic in cities has a significant effect on the road traffic congestions and as well the time it takes for road users to reach their destination. In this paper, we use the information of vehicles on the road network, including position, velocity, number of passengers, destination etc., as system inputs, to establish adaptive traffic signal coordination optimization model, in order to generate the optimal traffic signal area coordinated control scheme, and to suggest the best route to vehicles dynamically. We divide road network into grids, and use Mix-truncation-gauss-probabilitybased Interval Grey Number to describe vehicle position. The target of system optimization is to minimize the Total Trip Travel Time. To solve the problem, dynamic programming model is established, and the iterative algorithm is presented. A nice feature of our method is to recommend the shortest paths for vehicles when optimizing signal timing scheme of each intersection, which is called co-learning. Simulation results show that, the proposed method outperforms the Fixtiming method and Vehicle Actuated method on multiple evaluation indexes, including the Average Vehicle Delay Time, Average Vehicle Queue Length, Stops, Total Trip Travel Time and so on.