An Improved Traffic Signal Control Method Based on q-Learning

Abstract

The invention relates to an improved traffic signal control method based on Q learning. The method comprises the following steps that 1, initial lookup tables corresponding to all intersections inside a region are obtained under the simulation environment, and the initial lookup tables include environment information states, intersection actions and corresponding Q values; 2, environment information of all traffic intersections is detected continuously through sensors installed on the traffic intersections under the actual traffic environment; 3, a local intersection is selected inside the region, an intersection signal switching algorithm is adopted for judging whether signal switching is needed at the local intersection or not according to the environment information of the traffic intersections and the initial lookup tables, and a function is updated according to the Q values to update the initial lookup tables; 4, another intersection is selected, the third step is executed again, and finally signal control of all the intersections inside the region is completed. Compared with the prior art, the method has the advantages of considering the intersection linkage, achieving computation accurately and conveniently and the like.

Description

technical field [0001] The invention relates to the control of regional traffic signals, in particular to an improved traffic signal control method based on Q-learning. Background technique [0002] Since the 21st century, the traffic problem has become the main bottleneck restricting economic development. Traffic congestion and congestion have brought a huge impact on the global economy. It is urgent to solve the traffic congestion and congestion. In the United States, in 1984, the delay caused by traffic congestion and congestion was about 1.2 billion vehicle hours (veh·h), and the loss caused about 120 billion US dollars; according to an estimate in 2005, the annual traffic delay in the United States will reach 6.9 billion vehicles. · 7.3 billion gallons of fuel wasted by road traffic congestion in 2010; delays due to traffic congestion in the United States increased by 57% in 2010; additional fuel consumption due to traffic congestion is as high as $9 billion. In the UK...

AI Technical Summary

Problems solved by technology

[0005] However, the use of Q-learning algorithm for regional traffic control still has some shortcomings. For example, adjacent intersections in the area will affect each other, and there is linkage between intersections. In addition, the intersections in the area are different, and the maximum traffic volume between intersections is also different are different, so they need to be treated differently when controlling

Images