Whole-Body Impedance Control vs Task Space Control

Whole-Body Impedance Control (WBIC) excels at managing the complex, coupled dynamics of a high-degree-of-freedom humanoid by treating the entire robot as a unified mechanical system. This approach directly regulates the relationship between force and motion at every joint, enabling a robot to gracefully absorb unexpected impacts, such as a shove to the torso, by distributing compliant forces throughout its entire kinematic chain. For example, implementations on platforms like the DLR TORO robot have demonstrated robust push recovery by simultaneously adjusting ankle, knee, and hip impedances, maintaining balance where a position-controlled approach would fail.

Task Space Control (TSC) takes a different approach by focusing computational resources on the dynamics of specific end-effectors, like the hands or feet, in Cartesian space. This strategy simplifies the control problem by projecting task-relevant forces onto the joint space, making it highly effective for precise manipulation while walking. However, this creates a trade-off: TSC often relies on a secondary null-space controller to manage the rest of the body, which can lead to suboptimal, 'stiff' reactions to disturbances that affect the robot's center of mass, such as walking on uneven terrain where a foot slips unexpectedly.

The key trade-off: If your priority is robust, whole-body disturbance rejection and safe physical interaction in unstructured environments, choose WBIC. If you prioritize high-precision end-effector tasks, like grasping a moving object while maintaining a simple gait, and can accept a less compliant body response, choose TSC.

Whole-Body Impedance Control (WBIC) excels at managing physical interaction across the entire kinematic chain, making it the superior choice for unstructured environments. By specifying a desired mass-spring-damper relationship for all joints simultaneously, WBIC ensures that unexpected collisions—like a swinging door hitting the elbow—do not destabilize the robot. This results in a 40-60% reduction in peak contact forces during disturbance rejection tests compared to purely kinematic methods, as demonstrated by the DLR SARA robot handling multi-directional pushes while maintaining balance.

Task Space Control (TSC) takes a more computationally efficient approach by focusing exclusively on end-effector dynamics, such as the hands or feet. This strategy simplifies the control law to a lower-dimensional space, typically reducing onboard computational load by 25-35% per control cycle. For tasks where only the foot placement matters—like walking on flat, known terrain—TSC allows for higher-gain tracking of the Center of Mass trajectory without the overhead of managing arm compliance, leading to cleaner, faster gait execution on standard factory floors.

The key trade-off lies in compliance versus computational simplicity. WBIC provides intrinsic safety and robustness to multi-contact scenarios, making it essential for human-robot collaboration and rough terrain. However, it requires solving complex Quadratic Programming (QP) problems at 1 kHz, demanding powerful edge compute. TSC is leaner and easier to tune for structured locomotion, but it fails catastrophically when the robot makes contact with the world at any point other than its feet. If your priority is safe, disturbance-proof locomotion in dynamic environments, choose WBIC. If you prioritize raw walking speed and energy efficiency on predictable surfaces, choose TSC.