Studies on Q value-based Dynamic Programming with Boltzmann Distribution

Unlike conventional shortest route search algorithms, Q value-based Dynamic Programming with Boltzmann Distribution is proposed for providing reliable optimal candidate routes to avoid traffic congestions. Based on the multiple static traffic data, the characteristics of the method are systematically studied: 1) Candidate routes for each vehicle are generated probabilistically by Boltzmann Distribution policy; 2) Optimal temperature parameter is discussed considering different traffic volumes, which enables the method to adapt to the changing traffics. Bureau of Public Roads (BPR) volume-delay function is adopted to evaluate the traveling time of sections and help reassign the traffic volumes to sections in order to get reliable candidate routes. Numerical experiments are done to test the proposed method using Kitakyushu City´s map, comparing with conventional Q method. The results prove the effectiveness and efficiency of the proposed method, taking less average traveling time and providing more reliable optimal routes than Q method to avoid congestions in heavy traffics.