Safety in Admittance Control Using Reference Trajectory Shaping

This paper presents a switched model-reference admittance control framework designed for safe and compliant human–robot interaction via reference trajectory shaping. The proposed approach dynamically generates variable admittance parameters to satisfy task requirements while strictly maintaining task-space safety. To enhance robustness against unmodeled dynamics, a defined disturbance bound is integrated into the control law. Moreover, safety guarantees are explicitly formulated via invariance control, which limits the reference trajectory to the admissible region. A strict stability analysis for the switched system based on a common quadratic Lyapunov function guarantees the asymptotic convergence of the tracking error. The presented framework is verified for effectiveness through simulations on a two-link manipulator and shows significant performance gains over existing baseline methods. Finally, a real-time hardware implementation on a single-link manipulator is demonstrated to validate the controller’s practical feasibility and showcase its ability to ensure compliance and safety in physical interactions.