Automatic Guidance Of Cotton Bank Based On Reinforcement Learning Radial Planning

Manual operation of forklifts for loading and unloading goods is a common practice in cotton warehousing and logistics transportation.However,it has many drawbacks.For example,it increases the workload and risk of injury for workers,decreases operational efficiency and quality,and affects the safety and standardization of goods.In order to address these issues,many enterprises have started to use automated equipment to replace human labor for logistics operations,thereby improving efficiency,ensuring safety,and saving costs.In this article,a deep reinforcement learning algorithm is used to plan the path of a cotton shed guide car,and a simulation environment is built on the Robot Operating System(ROS)for experimental verification.The specific work is as follows:(1)Introduce a single guide car path planning method based on deep reinforcement learning-gray wolf algorithm(DQN-GWO).This method sets parameters such as state space,action space,reward function,initial parameters,convergence factor and weight strategy,builds a cotton warehouse environment model on grid map,and uses gray wolf algorithm to optimize the local optimal problem existing in network weight update process in deep reinforcement learning algorithm.By proposing the DQN-GWO algorithm,this article offers a new solution for logistics operations in cotton warehouses,which can effectively enhance the performance and efficiency of single AGV path planning.(2)A multi-Automated Guided Vehicle path planning method with improved MADDPG algorithm is introduced.This method takes into account that multiple groups of guided vehicles are often required to perform handling tasks together in cotton warehouses,and that there is mutual influence and competition between each group of guided vehicles.The algorithm framework in this study utilizes an adaptive gradient function and shared experience pool to describe interactions between multiple Automated Guided Vehicles,and an improved version of the DDPG algorithm to enhance the collaboration and action capabilities of such Automated Guided Vehicles.The MADDPG algorithm proposed herein effectively resolves multi-Automated Guided Vehicle path planning problems and offers a new coordination mechanism for logistics operations in cotton warehouses.(3)Analyze the results obtained by two algorithms when performing experimental verification in ROS simulation environment,and compare them with traditional methods.The experimental results show that both algorithms can effectively solve the problem of cotton warehouse guide car path planning,and are better than traditional methods in terms of running time,path length,collision times,etc.This paper realizes the application of two algorithms in cotton warehouse logistics operation through ROS simulation environment and agv model,laying a foundation for extending them to real scenarios in the future.In summary,this paper proposes two path planning methods for cotton depot guided vehicles based on deep reinforcement learning algorithms,which are respectively studied for single guided vehicle and multiple guided vehicles.The experimental results show that these two algorithms can effectively improve the efficiency and security of cotton warehouse logistics operations,and have a broad application prospect.