High-Performance Interaction Force Control: A Review of Actuator Impedance and Observer-Based Strategies

This paper presents a structured review of interaction force control in robotics, approached from the perspective of mechanical impedance and actuator characteristics. We classify actuators into three types, i.e., Type I: rigid actuators with direct force sensing, Type II: compliant actuators with elastic elements, and Type III: low-impedance actuators with model-based estimation. We further analyze how their inherent mechanical impedance constrains achievable performance. For each type, advanced observer-based control methodologies are introduced to mitigate limitations such as sensor drift, backlash, and model dependence. These include the Reduced-Order Multisensor-Based Force Observer (RMFOB) for drift compensation, the Transmission Force Observer (TFOB) for backlash-robust SEA torque estimation, and Frequency-Shaped Impedance Control (FSIC) for friction-robust model-based actuation. Experimental results across different actuator types verify improvements in transparency, stability, and impedance fidelity. In this review, we systematically link actuator properties with control strategies to provide a unified framework that bridges hardware constraints and control design. In the end, this offers practical guidance for next-generation applications in physical Human-Robot Interaction (pHRI), rehabilitation, and dynamic locomotion.