Research On Method Of Autonomous Driving Decision-making Based On LSTM And Grasshopper Optimization Algorithm

With the research of autonomous driving gradually moves towards practicality,people have higher and higher requirements for the accuracy and safety of autonomous driving.The self-driving system is a highly autonomous system,including perception module,path planning module,behavior decision-making module,and adaptive control module.The behavior decision-making module is the key technology to determine the safety and stability of autonomous vehicles.Main existing decision-making algorithms are divided into three categories,rule-based methods,reinforcement learning methods,and deep learning methods.Despite the high accuracy of simple scenarios in the rule-based methods,the complexity of the rule-making method limits its further development in complicated environments.Reinforcement learning methods made a safe and efficient driving decision in uncomplicated condition,the complexity of truly driving environments makes the accuracy of decision-making is limited.Therefore,deep learning methods are applied to autonomous driving decision-making,and long short-term memory network is one of the most important network architectures of decision-making for self-driving vehicles.However,the decision-making accuracy of the long short-term memory network is limited because it not consider the information of the surrounding vehicles,which is essential for the decision-making of autonomous vehicles.Secondly,softmax classifier is used in existing algorithms,which leads to their poor classification accuracy.So we use support vector machines for classification.Third,fixed parameters of kernel function lead to poor classification ability in support vector machine.In this paper,the decision-making method of autonomous driving is studied systematically,a novel architecture network is proposed for the above-mentioned problems.The network architecture is based on long and short-term memory neural network and support vector machine optimized by an improved grasshopper optimization algorithm for classification,called the autonomous driving decision-making algorithm based on long and short-term memory neural network and grasshopper optimization algorithm(GOA-Im LSTM).The following three innovations are mainly proposed in GOA-Im LSTM:(1)In order to consider information of surrounding vehicles,a novel network architecture is designed,used to extract vital features for self-driving vehicles,with three parallel long short-term memory network units and a long short-term memory network unit serial connected according to vehicle location is constructed.(2)In order to improve the classification accuracy,support vector machine with stronger classification capability than softmax is introduced to complete the classification task.(3)In order to improve the classification ability of the support vector machine,grasshopper optimization algorithm is used to optimize the parameters of support vector machine.Moreover,the exploration and exploitation capabilities of grasshopper optimization algorithm is uncoordinated.In this paper,dynamic weights in position movement formula are defined to balance the exploration and exploitation capabilities of grasshopper optimization algorithm.This paper experiments on Next Generation SIMulation(NGSIM)and compares the proposed algorithm with other autonomous driving decision-making algorithms.The experiment shows that GOA-Im LSTM has better autonomous driving decision-making results and it improves the accuracy of the autonomous driving decision-making.Autonomous driving decision-making is the focus of this article.During the experiment,this article proposes and solves the problems step by step.The above three improvements are proposed in order by gradually solving the above three problems.In addition,the algorithm in this paper can set different parameters according to specific experimental requirements and characteristics of dataset to adapt to different autonomous driving decision-making environments.The algorithm has strong flexibility.