Study On Traffic Light Scheduling For Traffic Flow Optimization In Entire Road Networks

With the continuous development of China’s urbanization, road traffic problem is becoming severe. Although traditional urban traffic light control system may be satisfy the need of city traffic system, as the number of urban vehicles increasing sharply, traffic jams has been seen frequently everyday. The obvious drawback of traditional traffic signal control system is that traffic light conversion time is fixed. And this system is hard to satisfy the requirement of city road traffic. For the sake of improving traffic flow fluency of entire road networks, there were many studies about intelligent traffic light scheduling algorithms (ITLS). Many researchers have considered that the traffic lights status are set according to the real-time traffic flows, which is used to optimize entire road networks traffic flows.In order to increase traffic efficiency all over the road networks and improve traffic capacity of each intersection, this paper studies ITLS algorithms based on Genetic Algorithm (GA) merging with Machine Learning (ML). The main works and achievements of this paper are as follows:1. This paper proposed an ITLS algorithm Based on GA (ITLSG). Traditional traffic signal control system switches over traffic lights by fixed time. However, it cannot adjust traffic signals display time dynamically by the shifty traffic flows. So, we proposed ITLSG which can search a best signals scheduling scheme to make entire road networks traffic flows achieve optimization based on real-time traffic flows. This algorithm is based on GA, and its characteristic is that it converts traffic lights state setting problem into searching the optimal solution problem by GA. It also puts encoding and decoding process into the traffic scheduling scenarios. Simulation results show that the performance of ITLSG is better than traditional traffic light scheduling algorithm.2. This paper proposed an ITLS algorithm Based on GA merging Linear Regression (LR) (ITLSGMLR). ITLSGMLR algorithm is improved based on ITLSG. Using linear regression to predict traffic flow of every road of each intersection, aimed to know the whole networks traffic flow state of each intersection of next unit time. The algorithm selects best traffic lights scheduling scheme that will tempt to optimize traffic flow at next unit time by GA’s fitness function. Real-time traffic flow prediction can help GA search a better scheduling solution. This can be used to optimize entire road networks traffic flows. Simulation results show that the performance of ITLSGMLR is better than ITLSG.3. This paper proposed an ITLS algorithm Based on GA merging Autoregressive model (ITLSGMAR). ITLSGMAR algorithm is improved based on ITLSGMLR. Due to unpredictable disruptions, such as accidents, road closure, the history traffic flow data may be cannot reflect the characteristics of traffic flows state. This will lead to inaccurate forecasts. In order to solve this problem, ITLSGMAR uses autoregressive model to improve prediction effect which is different from ITLSGMLR. This can help algorithm get more accurate information about traffic flow state and tempt to GA to search better traffic scheduling scheme. Simulation results show that the performance of ITLSGMAR is better than ITLSGMLR.