Reinforcement learning (RL) has been frequently used in recent years to develop intelligent robot agents that collaborate with human workers. In human-robot collaboration (HRC), the adaptation ability of automated robots is critical for efficient collaboration. However, previous studies using RL for assembly lines with HRC did not properly consider the possibilities of dynamics changes in the environment when collaborating with other human workers. In this study, we propose a method for adaptive assembly line balancing to efficiently collaborate with diverse workers having different task proficiencies through Bayesian RL. We mathematically formulate a Hidden-parameter Markov Decision Processes (HiP-MDP) model to represent the task proficiency of unknown workers in sequential assembly tasks. The robotic arm learns generalized policy to collaborate with diverse workers through meta-learning, and in a practical implementation, uses the policy adaptively by estimating the task proficiency of workers through Bayesian inference. We demonstrate the superiority of the proposed method in both virtual and real-world environments. Numerical studies show that the proposed method can collaborate well with diverse new workers.
18 April 2025 in International Journal of Machine Learning and Cybernetics
This study addresses the challenge of fleet design optimization in the context of heterogeneous multi-robot fleets, aiming to obtain feasible designs that balance performance and costs. In the domain of autonomous multi-robot exploration, reinforcement learning agents play a central role, offering adaptability to complex terrains and facilitating collaboration among robots. However, modifying the fleet composition results in changes in the learned behavior, and training multi-robot systems using multi-agent reinforcement learning is expensive. Therefore, an exhaustive evaluation of each potential fleet design is infeasible. To tackle these hurdles, we introduce Bayesian Optimization for Fleet Design (BOFD), a framework leveraging multi-objective Bayesian Optimization to explore fleets on the Pareto front of performance and cost while accounting for uncertainty in the design space. Moreover, we establish a sub-linear bound for cumulative regret, supporting BOFD's robustness and efficacy. Extensive benchmark experiments in synthetic and simulated environments demonstrate the superiority of our framework over state-of-the-art methods, achieving efficient fleet designs with minimal fleet evaluations.
29 October 2026 in Simulation Modelling Practice and Theory