An extended flexible job scheduling problem is presented with characteristics of technology and path flexibility (dual flexibility), varied transportation time, and an uncertain environment. The scheduling can greatly increase efficiency and security in complex scenarios, e.g., distributed vehicle manufacturing, and multiple aircraft maintenance. However, optimizing the scheduling puts forward higher requirements on accuracy, real time, and generalization, while subject to the curse of dimension and usually incomplete information. Various coupling relations among operations, stations, and resources aggravate the problem. To deal with the above challenges, we propose a multi-agent reinforcement learning algorithm where the scheduling environment is modeled as a decentralized partially observable Markov decision process. Each job is regarded as an agent that decides the next triplet, i.e., operation, station, and employed resource. This paper is novel in addressing the flexible job shop scheduling problem with dual flexibility and varied transportation time in consideration and proposing a double Q-value mixing (DQMIX) optimization algorithm under a multi-agent reinforcement learning framework. The experiments of our case study show that the DQMIX algorithm outperforms existing multi-agent reinforcement learning algorithms in terms of solution accuracy, stability, and generalization. In addition, it achieves better solution quality for larger-scale cases than traditional intelligent optimization algorithms.
4 October 2022 in IEEE Journal of Radio Frequency Identification
Reinforcement learning is emerging to achieve real-time response and near-optimization for solving the flexible job shop scheduling problem (FJSP), an important and NP-hard problem for intelligent manufacturing systems. Although some methods based on reinforcement learning have been proposed to solve the FJSP, theres still room for improvement. In this paper, we propose a new approach called reinforcement learning with the generative adversarial network (RLGAN) based on the parallel learning framework. A simulation-based artificial workshop system is established to generate a large number of sample plans as a training set for RLGAN to develop a near-optimal scheduling model. The case study shows that an implementation of our proposed method, QTRAN-GAN, can generate near-optimal plans and outperforms the corresponding pure reinforcement learning method, QTRAN.
10 February 2026 in IEEE Transactions on Automation Science and Engineering
This study introduces iProDMP, a novel framework that integrates human behavioral preferences into exoskeleton learning through unified probabilistic movement modeling. The framework addresses three critical limitations: Dynamic movement primitives (DMPs)’ inability to represent preference uncertainty, inadequate trajectory generation beyond observation range in probabilistic movement primitives (ProMPs), and unreliable via-point modulation in hybrid approaches. Our solution features three innovations: First, we establish dynamic-probabilistic consistency conditions for unified DMP-ProMP frameworks, which enable stochastic modeling of human preferences while preserving attractor stability. Second, a novel scaling method decouples shape modulation from model hyperparameters, enabling flexible motion adaptation. Third, consistency-guaranteed expectation-maximization resolves parameter optimization within the unified framework. Experiments on lifting trajectory imitation demonstrate strong extrapolation beyond demonstration distributions, particularly for via-points outside training data, thereby validating adaptability to variable task conditions. In hip exoskeleton-assisted lifting tasks, our approach achieves a 28% improvement in assistance efficiency over conventional implementations.
1 January 2026 in IEEE Transactions on Industrial Electronics