|
|
|
| Adaptive Robot Scheduling Using SP–MCTS for Industrial Internet of ThingsEnabled Hybrid Flow Shop |
| GUO Jian 1,2，SHI Yaoyao 1,2，HU Hao1 ，CHEN Zhen 1,2，ZHANG Junfeng3 ，ZHAO Pan 1,2 |
| (1. School of Mechanical Engineering, Northwestern Polytechnical University, Xi’an 710072, China; 2. Key Laboratory of High Performance Manufacturing for Aero Engine, Ministry of Industry and Information Technology, Northwestern Polytechnical University, Xi’an 710072, China; 3. School of Mechanical and Electrical Engineering, Xi’an University of Architecture and Technology, Xi’an 710055, China) |
|
|
|
|
Abstract With the deep integration of industrial internet of things (IIoT) and artificial intelligence (AI) technology, automated guided vehicles (AGVs) and mobile robots have been widely used in internet of things-enabled floor shop. In view of many complex factors such as real-time dynamic changes and uncertain conditions in the workshop, the SP– MCTS (Single-player Monte-Carlo tree search algorithm) method with each job group as a subtree and real-time state as the root node is proposed to implement adaptive scheduling of workshop. The problem of robot scheduling is formulated as a Markov decision process (MDP) in which state representation, action representation, reward function, and optimal policy, are described in detail. In the real-time scheduling process, the proposed search method establishes a subtree for each job group, and only the state relationship between two adjacent groups is considered in optimization, thereby the calculation difficulty is simplified. In the subtree search process, SP–MCTS is used to search with the real-time state as the root node. At the same time, the expansion method and the pruning method are used to carry out strategy exploration and information accumulation respectively. Therefore, the deeper the real-time status node in the subtree, the faster and more accurate the optimal strategy is obtained. The case study based on a real-world shop is proven and the results validate the effectiveness and superiority of the proposed approach.
|
|
|
|
|
|
|
|
|
2021, 
