Deep reinforcement learning based strategy for optimizing phase splits in traffic signal control
2022 IEEE 25th International Conference on Intelligent …, 2022•ieeexplore.ieee.org
Urban traffic networks are often choked due to recurrent congestion. Heavy economic costs,
environmental pollution and severe noise pollution can arise from the lack of valid Traffic
Signal Control (TSC) strategy. In this paper, assuming the cycle time of traffic lights and the
phase order of a signal cycle are fixed, a Deep Reinforcement Learning (DRL) algorithm
called Twin Delayed Deep Deterministic Policy Gradient (TD3) is first investigated to control
the traffic signal by optimizing phase splits. Unlike widely used Deep Q-learning based TSC …
environmental pollution and severe noise pollution can arise from the lack of valid Traffic
Signal Control (TSC) strategy. In this paper, assuming the cycle time of traffic lights and the
phase order of a signal cycle are fixed, a Deep Reinforcement Learning (DRL) algorithm
called Twin Delayed Deep Deterministic Policy Gradient (TD3) is first investigated to control
the traffic signal by optimizing phase splits. Unlike widely used Deep Q-learning based TSC …
Urban traffic networks are often choked due to recurrent congestion. Heavy economic costs, environmental pollution and severe noise pollution can arise from the lack of valid Traffic Signal Control (TSC) strategy. In this paper, assuming the cycle time of traffic lights and the phase order of a signal cycle are fixed, a Deep Reinforcement Learning (DRL) algorithm called Twin Delayed Deep Deterministic Policy Gradient (TD3) is first investigated to control the traffic signal by optimizing phase splits. Unlike widely used Deep Q-learning based TSC strategies, the action space of TD3 is continuous so that a determined phase split can be obtained at each time step. The Markov Decision Process is properly formulated by putting forward innovative representations of state, action, and reward. Several implemented experiments have well showcased the high control performance of the proposed TD3-based TSC strategy via a microscopic traffic simulation platform.
ieeexplore.ieee.org
Showing the best result for this search. See all results