Design And Implementation Of Robot Shelf Picking System Based On ROS

With the promotion of intelligent manufacturing strategy,intelligent logistics and the storage system have become the new development direction of manufacturing logistics.At present,most of the business of warehouse logistics has been automated.However,due to the complexity of the tasks in the shelf picking process,most of the goods are still carried out manually,accounting for more than 60% of the workload of the entire logistics process.Picking methods and technologies directly affect the picking efficiency of the entire warehouse and are related to the level of warehouse costs,so they have attracted much attention.The rapid development of the Robot Operating System(ROS)and the digitization of the production process provide a new opportunity for the realization of fully automated warehouse picking in warehouses.Based on the application requirements and technical difficulties of automated shelf picking,this thesis designs and implements a robot automatic shelf picking system based on ROS.The system consists of two parts,a robot picking control system based on ROS and a robot picking planning system based on digital twins.The robot picking control system is mainly used for shelf environment sensing and picking task execution and runs in the physical production environment.To improve the reliability of the control system and reduce the cost of research and development,this thesis uses ROS with a simplified architecture and rich functions as the basic software framework.In response to the needs of item detection and positioning in the shelf picking task,the system integrates multi-sensor sensing technology to improve the detection accuracy.The robot picking planning system is mainly used for picking strategy planning and simulation verification.To solve the problem that most of the current robot picking systems are difficult to achieve rapid verification and deployment of online optimization algorithms,this thesis introduces digital twin technology.The ROS interface is used to complete the quasi-real-time data interaction between the physical world and the digital world,enabling online planning and rapid verification of dynamic physical environments.To optimize the planning results,this thesis proposes an order picking strategy planning method based on the improved Nondominated Sorting Genetic Algorithm ?(NSGA?)algorithm.The order-picking sequence is planned with the shortest picking path and the highest executable.The robot control system and the robot picking planning system communicate in real-time,coordinate and cooperate to form a robot shelf picking system with a complete physical-digital-physical closed-loop.The thesis validates the feasibility and effectiveness of the robot picking system based on ROS through the object picking experiment of the robot control system,the performance comparison experiment of the picking planning system,and the robot shelf picking system comprehensive experiment.