10 October 2024 in Optics & Laser Technology
11 August 2025 in IEEE Robotics and Automation Letters
Automatic assembly of board-to-board (BTB) connectors remains a significant challenge in smartphone manufacturing due to severe visual occlusion, tight assembly tolerances, and process constraints that prohibit separate visual adjustment stations. This letter proposes Visual-Privileged Co-Learning (VPCL), a novel privileged reinforcement learning (RL) algorithm specifically tailored for robust, force-guided precision assembly in partially observable environments. Leveraging additional visual privileged information only during training and relying solely on force feedback during deployment, VPCL employs a dual-actor (teacher-student) architecture within the Soft Actor-Critic (SAC) framework. To enhance teacher-student interaction in the co-learning process, VPCL introduces a Teacher Quality Loss to address modality imbalance and a probabilistic Policy Selector to mitigate data distribution shifts. Extensive comparative and ablation experiments in real-world smartphone assembly tasks demonstrate that VPCL achieves superior performance and robustness, as well as a remarkable zero-shot transfer capability, compared to existing baselines.
2 March 2026 in IEEE Transactions on Automation Science and Engineering
In many high-precision industrial applications, robots are deployed to perform precision peg-in-hole assembly on large batches of manufactured pegs and holes. When the nominal design adopts a transition fit, machining errors may cause each peg–hole pair to exhibit either a clearance fit or a slight interference fit, with an unknown and continuously varying fit amount. This paper addresses robotic batch precision assembly tasks under uncertain fit types and fit amounts, and proposes a novel multi-stage framework for systematically constructing a robust assembly control strategy. The overall batch precision assembly task is first decomposed into multiple deterministic subtasks characterized by different but fixed fit amounts. For these subtasks, a force-vision fusion controller-driven reinforcement learning method combined with a multi-task reinforcement learning training framework (FVFC-MTRL) is developed to jointly learn multiple compliance control strategies. Subsequently, a multi-teacher policy distillation scheme is designed to integrate multiple trained strategies into a single student strategy, yielding a robust control strategy. Real-world experiments demonstrate that the proposed method successfully constructs a robust control strategy for high-precision assembly tasks with varying fit types and fit amounts. Moreover, the MTRL framework significantly improves training efficiency by 50%, and the final control strategy achieves superior force compliance and a higher success rate compared with several existing methods. Note to Practitioners—Unlike most existing studies that learn an assembly strategy for a specific peg–hole pair with a known clearance fit, this work addresses the challenges of control robustness and compliance in robotic batch precision assembly with uncertain fit types and fit amounts, which is closer to practical industrial production settings. Reinforcement learning is combined with a pre-designed force-vision fusion controller to enhance both training stability and control performance. Furthermore, a multi-stage framework, consisting of task decomposition, multi-task reinforcement learning, and policy distillation is developed to endow the control policy with robustness. The final constructed control strategy can be directly deployed in industrial batch precision assembly tasks.