Torquecontrolled

Torque-controlled refers to a mode of operation in robotic actuators and control systems in which joint torque commands are generated and applied directly, rather than commanding joint positions or velocities. In torque control, the control loop aims to ensure the actual joint torque matches a desired torque, often using feedback from torque sensors or estimation from motor current and a dynamic model.

Implementation typically uses a cascaded control structure: an inner torque or current control loop inside an outer impedance or position loop. The outer loop may specify a desired interaction behavior, such as stiffness and damping, enabling compliant interaction with humans or the environment. System design requires accurate torque sensing or high-fidelity torque estimation, fast control bandwidth, and accurate actuator models to translate torque commands into actuator torque.

Advantages include improved contact safety and compliance, better force regulation, and robust performance under unknown contact conditions. Torque control can facilitate compliant tasks, haptic feedback, and teleoperation by providing more direct control of interaction forces and torques.

Challenges involve higher hardware requirements, sensor noise, friction, backlash and model inaccuracies that can hamper torque tracking. Ensuring stability in interaction with stiff environments can be nontrivial, and torque control demands high sampling rates and low-latency actuators. Hardware may require integrated torque sensing, precise current sensing, or sophisticated estimators to maintain accurate torque feedback.

Applications span variable-stiffness or collaboration robots, prosthetic limbs and exoskeletons, haptic devices, and robot teleoperation. It is often contrasted with position control, which tracks a target position, and with impedance or force control that shapes interaction behavior through indirect means.