Researchers have proposed a control strategy that integrates admittance control with reinforcement learning to address key challenges in human-robot interaction (HRI), such as insufficient compliance, high collision risk, and poor tracking performance. The method generates a smooth, differentiable reference trajectory using a soft saturation function and an expected admittance model, ensuring the end-effector remains within a predefined spatial boundary.
Key takeaways
- Reinforcement learning with an actor-critic structure compensates for system dynamic uncertainties, improving robustness in complex tasks.
- A time-varying asymmetric barrier Lyapunov function enforces strict position constraints, enhancing safety during HRI.
- Lyapunov stability analysis proves semi-global uniform ultimate boundedness of the closed-loop system.
- Experiments on a Baxter robot show superior tracking accuracy, compliance, and collision avoidance compared to adaptive impedance, fuzzy adaptive impedance, and traditional impedance control.
This approach offers a practical solution for collaborative robots operating in constrained spaces, enabling precise trajectory tracking while safely adapting to human-applied forces. It is particularly relevant for manufacturing automation where human-robot collaboration requires both flexibility and strict safety guarantees.
Source: 《机器人》期刊 (robot.sia.cn) · Published 2026-08-24 · “面向人机交互的导纳与强化学习融合控制策略”