Research On Dynamic Path Planning For Intelligent Transportation

Vehicle path planning is one of the important research directions in intelligent transportation,and its research content mainly includes road network models,traffic information prediction and path planning algorithms.How to make full use of road traffic information and dynamically and quickly provide a reasonable and efficient driving route for the target vehicle is of great significance.The thesis integrates deep reinforcement learning into the path planning algorithm,and studies the problem of vehicle dynamic path planning.The main work of the thesis is as follows:Aiming at the problem of incomplete analysis of road conditions in existing path planning,the thesis constructs a road efficiency index evaluation model based on the analytic hierarchy process,comprehensively analyzes various static and dynamic indicators that affect road traffic efficiency,and obtains the estimated traffic cost of each road section.Aiming at the path planning problem under the constraint of passing points,the thesis combines the simulated annealing method to study a global path planning algorithm,which selects a path with a lower traffic cost and improves the search efficiency of the algorithm.Through the simulation in the urban road network model,the running time of the the thesis’ s algorithm is reduced by 77% and 19% respectively compared with the two comparison algorithms.Finally,the the thesis build an intelligent transportation route planning demonstration platform to obtain real-time road condition information and visually display dynamic route planning related functions.Based on the above algorithm,the the thesis studies a dynamic path re-planning algorithm based on deep reinforcement learning for the problem that the subsequent driving path needs to be dynamically adjusted according to changes in traffic information during the driving process of the vehicle.The algorithm uses the deep Q network as the basic framework to dynamically perceive the target vehicle and road network information,and at the same time,through continuous learning of changes in the road network information,it estimates the changes in traffic flow,and uses the path re-planning model to evaluate and make decisions on the current path to make the overall situation.Re-plan the path to obtain a better route.The thesis builds different simulation scenarios to evaluate the performance of the algorithm through SUMO,and compares it with the traditional algorithm.The experimental results show that the algorithm of the thesis reduces the average journey time of the target vehicle by 36.7% in the urban road network environment,and at the same time,it has been optimized in terms of the proportion of waiting time and average driving speed.