Research On The Path Planning Of Unmanned Vehicles Logistic Distribution

With the rapid development of our country’s logistics,the volume of express delivery business has exploded,the cost of manual distribution is getting higher and higher,and the society’s requirements for distribution efficiency are getting higher and higher.The traditional logistics based on manual distribution can no longer meet the needs of society.In recent years,the rise of unmanned driving technology has provided a feasible solution for logistics distribution.Path planning is one of the main problems of logistics distribution.Research on unmanned vehicle path planning technology assists unmanned vehicles in intelligent route planning,which can improve logistics Distribution efficiency is of great significance.This paper separately researches the global path planning and local path planning of unmanned vehicle logistics distribution.The main research contents are as follows:(1)For this problem of lag and poor real-time performance in global route planning for unmanned vehicle distribution,which leads to unreasonable path planning,a global route planning model for unmanned vehicle distribution based on traffic situation awareness is proposed.In order to obtain real-time traffic situation information,a traffic situation crawler framework is built based on python;in order to predict traffic flow information,a traffic flow prediction model is built based on the GRU network;In order to quantitatively evaluate the road network,a traffic situation weighted fusion model is designed,which fuses the real-time traffic situation information and the predicted traffic flow,obtains the road network model with traffic situation cognition,and then carries out the global path planning.(2)For this problem of poor local autonomous planning ability for unmanned vehicle distribution and difficulty in adapting to the intricate traffic environment,a local path planning model based on heuristic reward and adaptive exploration strategy DQN is proposed.In order to solve the problem of sparse reward in path planning of ordinary DQN,a heuristic continuous reward function is designed to make each state of the agent have comprehensive reward feedback in the learning process.In order to balance the problems of exploration and utilization in reinforcement learning,an adaptive exploration strategy is proposed to enable agents to rationally utilize "old knowledge" and explore "new space".Finally,the improved DQN path planning model is integrated with heuristic reward and adaptive exploration strategy.(3)Perform simulation experiments and analysis on the proposed models respectively.Taking the rectangular area from Beijing’s Jimen Bridge to Dongshitiao Bridge as the experimental area,experiments are carried out on the road network with and without traffic situation information based on the ArcGIS platform.The experiment shows that the road network path planning based on traffic situation cognition is better than the shortest path planning,which can effectively avoid the congested road sections,save the delivery time and improve the delivery efficiency.A simulation environment is built based on the python-based tensorflow and keras modules,and the improved DQN ablation experiment is performed.The experiment shows that the DQN model that combines heuristic rewards and adaptive exploration not only has high efficiency in path planning,but also the planning path is the most optimal.Global path planning based on traffic situation recognition can plan a more reasonable global path,save time,and improve distribution efficiency.When the unmanned distribution vehicle is in an unknown environment,the local path planning with improved DQN is used to adjust the global route in real time.The path planning combines with the global and the local makes the unmanned vehicle logistics distribution route planning more rational and "intelligent",and to a certain extent can speed up the application of unmanned vehicle distribution.