Signal Timing Optimization Based On Short-term Traffic Flow Prediction

In the current era of increasing demand for convenient travel,intelligent transportation systems are playing an extremely important role,and relying solely on traffic police or monitoring methods can no longer meet various travel needs.How to predict road traffic professionally and accurately is of great significance to timely improve traffic conditions and ease traffic congestion.At the same time,the optimization of signal timing based on short-term traffic flow prediction is also an important means for the construction of intelligent transportation systems.This paper uses the law of traffic flow conservation and machine learning to implement a short-term traffic flow prediction method.At the same time,the signal timing scheme is linked to the traffic flow in real time through an optimization method,which improves the traffic efficiency of vehicles and alleviates traffic congestion.The main research work is summarized as follows:(1)In view of the existing traffic flow prediction models,which are based on existing data,they do not care about the internal rules of the data,and feature selection is mostly based on manual experience selection.This paper proposes multi-section short-term traffic flow prediction based on K nearest neighbor regression.The model uses real data as a data set,and uses parameters such as a signal timing scheme,upstream pavement traffic at a certain period,and downstream pavement traffic at a certain period as model features.A multi-segment short-term traffic flow prediction model based on K nearest neighbor regression is constructed.This method can improve the prediction accuracy,generalization ability and interpretability of the model,and achieve the purpose of high-precision traffic prediction.The example verification results show that the short-term traffic flow prediction model constructed has higher prediction accuracy.The prediction results show that the model has a root mean square error(RMSE)of 7.08 on the test data.Compared with the existing short-term traffic flow prediction methods,it has the characteristics of better accuracy,faster prediction speed and stronger practicability,and can perform efficient short-term traffic flow prediction.(2)Aiming at the problems of low accuracy,large objective function error and poor interpretability of existing optimization schemes of signal timing,a method of signal timing optimization based on improved Webster function and improved genetic algorithm was proposed.First,the method takes the traffic flow,the signal cycle and the green signal ratio as model inputs,the average vehicle delay at the intersection as the model output,and then uses the sum of the average vehicle delay in each direction and the product of the traffic flow in each direction to build the total vehicle delay function at the intersection and use this as the objective optimization function.This method can realize the transformation from multi-objective optimization function to single-objective optimization function;Secondly,after completing the construction of the objective optimization function,in order to solve the objective function with multiple constraints,this paper introduces an improved genetic algorithm to minimize the total delay at the intersection by changing the signal timing;Finally,the intersection delay of the optimized signal timing scheme is compared with that of the original signal timing scheme to verify the feasibility of the scheme.Compared with existing methods,this method has a higher precision objective function,and can solve optimization problems with multiple constraints during the solution process.The experimental results show that the semaphore timing scheme optimized by the improved genetic algorithm can reduce the intersection delay under the original timing scheme by nearly 15.37%.(3)Aiming at the problem of how to change the method of signal timing from a single signal optimization to the coordinated optimization of adjacent signals and the slow solution of the signal timing scheme,an optimization model of “optimized model of adjacent intersection signal lights based on predicted traffic flow and XGBoost" was proposed.In order to analyze the relationship between "semaphores-semaphores",the original "semaphores-semaphores" mode was changed to "semaphores-segments-semaphores" mode,and the impact of the timing of semaphores on road traffic was tapped.Afterwards,optimize the traffic lights at adjacent intersections based on the predicted traffic flow at adjacent intersections.In addition,in order to solve the problem that the convergence speed of the improved genetic algorithm is too slow,under the premise of ensuring the accuracy of the solution,the XGBoost model is introduced to greatly improve the response speed of the signal timing solution process.The experimental results show that the “adjacent intersection signal light optimization method based on predicted traffic flow and XGBoost” can reduce the delay of adjacent intersections under the original timing scheme by nearly 22.75%,and the speed of signal timing is greatly improved.Finally,this paper comprehensively compares the advantages and disadvantages of the above methods from the theory,applicable conditions and scope,performance,and prediction errors,analyzes the shortcomings of the current research,and proposes directions to further improve the accuracy and efficiency of the model solution.It is expected that the accuracy of the model solution will be more accurately evaluated from the measured data.