A Decision-making Algorithm Of Autonomous Vehicle Based On Improved SVM

Due to the improvement of artificial intelligence theory,autonomous vehicle related technology has also been rapid development.Driverless driving is a comprehensive system with many functions,such as environment perception,decision-making,adaptive control and path planning.Among them,decision-making is like the brain of autonomous vehicle,which provides operation instructions for the vehicle to drive safely and smoothly.This is very important for autonomous vehicle.However,due to the complexity of the real traffic environment,the decision-making module of decision-making has not yet reached many requirements such as safety,reliability and timeliness.Once it is applied to the real road scene,it will cause immeasurable consequences.Therefore,it is necessary to carry out explore deeply and research on this module.Decision-making is essentially a classification problem by analyzing the surrounding environment information of vehicles and making corresponding driving decisions.In this paper,after deeply studying the background,significance and research status at home and abroad,it is found that SVM has good robustness and generalization,and can effectively solve the nonlinear classification problem.However,the classification effect of SVM is affected by the parameters.Therefore,this paper uses Bat optimization algorithm for optimization.Bat algorithm is a stochastic global optimization model,which can make use of the advantages of the population and integrate the global and local search to achieve the optimization effect.But,with the optimization iteration,the diversity of the population will decrease and fall into local optimum.To solve the above problems,this paper proposes a decision-making algorithm of autonomous vehicle based on improved Support Vector Machine:(1)When processing nonlinear data and outliers,Support Vector Machine will introduce parameters to be defined.Different parameter values will have a greater impact on the model.In order to create the optimal Support Vector Machine model,this paper proposes an improved bat algorithm to optimize it and return the optimal solution to improve the accuracy of the model.(2)In the process of bat algorithm optimization,all bat individuals always learn from the optimal bat,which leads to the gradual decline of diversity and easy to fall into the local optimal solution.In order to solve this problem,Levy flight is added to increase the diversity of bat individual position and expand the search range;in addition,adaptive frequency is adopted to adjust the bat individual position,which greatly improves the bat learning efficiency.(3)Due to the different needs in real life,there are vehicles with different functions and sizes.Different types of vehicles have different driving styles.This paper will use SVM model for quantitative analysis and verification.Through the verification of the model,it is found that large cars prefer a stable driving environment and will not change lanes easily,while small cars prefer a better driving environment and have more flexibility.Finally,this paper uses the next generation simulation(NGSIM)to verify the feasibility of the algorithm,and designs horizontal and vertical comparative experiments to verify the innovation of this paper one by one.The experimental results show that the algorithm has the highest accuracy rate of decision-making results.