Deep reinforcement learning traffic signal control method combined with state prediction

Abstract

The invention discloses a deep reinforcement learning traffic signal control method combined with state prediction. The method comprises the following steps: (1) modeling road network environment and traffic flow data; (2) selecting a deep reinforcement learning algorithm and designing three elements; (3) predicting a future traffic state; (4) training the model; and (5) carrying out experiment for test. The waiting time of vehicles can be shortened, and the passing efficiency of a road network is improved.

Description

technical field [0001] The invention relates to the technical field of intelligent traffic signal control, in particular to a deep reinforcement learning traffic signal control method combined with state prediction. Background technique [0002] With the improvement of living standards, the number of car ownership continues to increase, and the problem of urban traffic congestion is also becoming more and more serious. Traffic signal control is the most direct and cheapest way to improve road traffic efficiency and alleviate traffic congestion. Traditional signal control methods mainly include fixed timing control, sensory control and adaptive control. SCATS (Sydney Coordinated Adaptive Traffic System) and SCOOT (Split Cycle Offset Optimizing Technique) are currently widely used adaptive traffic signal control systems, which use simplified traffic models to solve optimal signal control strategies; however, the establishment of simplified models relies on a large number of B...

AI Technical Summary

Problems solved by technology

However, for the complex and changeable traffic flow in the actual scene, the optimal control strategy can only be obtained by integrating the current, historical and future states

[0004]Real traffic flow data has the characteristics of sudden change, real-time, periodicity, etc., and is a typical time series data. At present, the signal control methods based on DRL only use the current traffic flow State-based decision-making, limited control effect

Images