Treffer: Research on simulation teaching of multi-warehouse robot path planning in smart factories.

[Objective] The Smart Factory curriculum integrates disciplines such as computer science, automation, and information technology, emphasizing hands-on practice and skill development. However, traditional teaching methods often focus on theoretical knowledge regarding interactivity and participation and fail to promptly incorporate teaching content related to new technologies and equipment. This course develops a practical teaching strategy that encourages students to participate in factory demand research. By simulating real factory environments and production processes, this method introduces the latest technological achievements, namely the multi-agent reinforcement learning task supervisor, into smart warehouse system case studies, thereby enhancing students' skills and innovation. [Methods] To address challenges in path planning and real-time delivery sequence adjustment for multi-load warehouse robots operating in dynamic environments, we propose a comprehensive policy that integrates local path planning, dynamic obstacle avoidance, and real-time delivery cost evaluation. This paper models a warehouse environment and captures its layout, shelf positions, aisle widths, and potential obstacle locations to frame multi-objective delivery as a traveling salesman problem. To ensure safe navigation, the robot tasks were divided into movement, obstacle avoidance, and collision avoidance. In complex warehouse environments, robots frequently execute multiple behaviors concurrently. We introduce a null-space behavioral control algorithm to manage conflicts, where behaviors are projected into null spaces based on their assigned priorities to form composite behaviors. Priorities are determined by a multi-agent reinforcement learning task supervisor, who uses composite behavior velocities as actions for the deep reinforcement learning algorithm and the robot positions as states. Through continuous interaction with the environment during the learning process and offline training, the supervisor develops a priority selection policy. Furthermore, the delivery sequence for warehouse robots is dynamically adjusted using a scoring evaluation mechanism. This mechanism updates the selection of target delivery points in real time to minimize transportation costs. This policy ensures that each warehouse robot safely delivers multiple items over the shortest possible distance while avoiding collisions, thereby providing an optimal solution for commodity delivery in warehouse systems. [Results] This simulation compared the effectiveness of a reinforcement learning task supervisor to a multi-agent reinforcement version in a warehouse environment. Although warehouse robots equipped with a reinforcement learning task supervisor can navigate paths and avoid obstacles, frequent priority switches result in longer paths. By contrast, warehouse robots utilizing a multi-agent reinforcement learning task supervisor, enhanced by a scoring evaluation mechanism, can intelligently adjust the delivery sequence based on path length costs and optimize path selection through goal-oriented learning. This approach not only reduces transportation path lengths but also effectively shortens dynamic obstacle avoidance distances and waiting times due to refined obstacle and collision avoidance behavior designs. More stable priority switching significantly enhances delivery flexibility, decreases path lengths, and alleviates traffic congestion. [Conclusions] Considering the course characteristics and teaching challenges of smart factories, this paper presents a simulation teaching scheme for multi-warehouse robot path planning tailored to smart factories. It addresses challenges in multi-objective delivery and multi-task conflicts by using a scoring evaluation mechanism to dynamically update delivery sequences of target points, optimize robot transportation paths, and improve the overall efficiency of warehouse system operations. This teaching scheme leverages Python software platforms for model experiments, enhances practical teaching resources, and employs virtual simulations to compensate for hardware deficiencies. By evaluating different solutions, students gain a deeper understanding of smart factory requirements and enhance their problem-solving skills. [ABSTRACT FROM AUTHOR]

Copyright of Experimental Technology & Management is the property of Experimental Technology & Management Editorial Office and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)