Abstract: Factors affecting public transit system were synthetically analyzed. An innovative transit signal priority (TSP) strategy based on reinforcement learning algorithm was proposed. The trial and error mechanism of reinforcement learning were utilized, so the signal plans could be optimized iteratively by implementing them and estimating the rewards. The proposed idea made the TSP strategy have a capability of self-learning. Based on the software of Paramics, simulations were carried out. And the results demonstrated that the proposed TSP strategy could not only improve the efficiency of transit operation, but also reduce the impacts on general traffic at signalized intersections.