Mode-selective adaptive admittance force-position control for safe robotic interaction in unknown time-varying environments.
With the increasing complexity of robotic interaction tasks, robotic manipulators are required to safely interact with unknown and time-varying environments while maintaining accurate trajectory tracking. Such tasks involve frequent switching between free-motion and contact-motion phases, uncertain environmental parameters, and transient impact-induced force oscillations. To address these issues, this paper proposes a mode-selective adaptive admittance force-position control framework in task space. First, an activation matrix is introduced to select the force-controlled and position-controlled subspaces according to contact and tracking conditions. Second, an environment-aware adaptive admittance outer loop is developed by integrating Gaussian Process Regression-enhanced Extended Kalman Filter (GPR-EKF) estimation, Lyapunov-based parameter adaptation, and radial basis function neural network (RBFNN) compensation. Third, a mode-gated self-tuning PID inner loop is employed to improve trajectory tracking accuracy under different interaction modes. Simulation and experimental results demonstrate that the proposed method can reduce force overshoot, improve steady-state force tracking accuracy, and enhance safe interaction in unknown time-varying environments.