An adaptive timing mechanism for urban traffic pre-signal based on hybrid exploration strategy to improve double deep Q network

Abstract With the increasing traffic congestion in cities, priority of public transit has become a consensus for development and management urban transportation. The pre-signal mechanism, which gives time space to buses by optimizing road right-of-way allocation, gained wide attention application. In order broaden action exploration range agent avoid decision from falling into suboptimal strategy or local optimal strategy. For DDQN algorithm, this paper reduces probability directly selecting increases non-greedy actions based on principle that “the with larger value function is more likely be selected.” This addresses problem existing mechanism cannot adaptively adjust advance time, proposes adaptive timing Hybrid Exploration Strategy Double Deep Q Network (HES-DDQN) combining $$\epsilon $$ ϵ -greedy Boltzmann We have used simulation software VISSIM conduct experiments an intersection. experimental results show that, compared method setting no formula pre-signal, HES-DDQN can significantly reduce average delay buses, waiting queue length, number stops social vehicles.