Robot control
Self-collision avoidance
Definition
Self-collision avoidance keeps different parts of the same robot from intersecting or entering an unsafe clearance while it moves. It can be enforced during planning, inverse kinematics, or reactive control using the robot's geometry and current state.
Also known as: Robot self-collision avoidance, Self collision avoidance
Updated
Check the robot against itself
A robot model contains link pairs that can collide, such as a humanoid hand and torso or the two legs during a crossing step. Adjacent links that meet at a joint are usually exempted, while other pairs are checked using meshes or simpler shapes. The controller or planner maintains a nonnegative clearance, not merely a configuration that is collision-free at one instant.
This is different from environmental collision avoidance. Both use geometric distance and predicted motion, but self-collision geometry moves with the robot's own joints and can create many coupled pairwise constraints.
Planning and reactive enforcement
A motion planner can reject configurations or trajectory segments that intersect. An inverse-kinematics or whole-body controller can instead add distance inequalities, projected gradients, artificial potentials, or a control barrier function that modifies commands near a boundary.
Khazoom and colleagues combine barrier constraints with a whole-body controller for the MIT Humanoid. Their simulations include hand reaching and walking cases where the controller changes infeasible reference motion to avoid body contact. The paper explicitly evaluates a simulated robot; it does not report a hardware safety certification.
Clearance calculations have blind spots
Simplified spheres, capsules, or convex shapes can make distance queries fast but omit fine geometry. Discrete checks can miss a collision between sampled times, while continuous checks cost more. Flexible covers, cables, carried objects, clothing, and calibration error may not be present in the model.
Reactive avoidance also needs room and actuator authority. If a requested pose is already too close, or balance and joint limits leave no feasible escape direction, changing one task command may not prevent contact. Safety margins, prediction horizon, update rate, braking behavior, and state uncertainty all affect the result. A reported minimum model distance is not automatically the clearance of the physical machine.
Sources
Related terms
Motion planning
Motion planning finds a robot movement from an initial state to a goal while satisfying constraints such as collision avoidance. A planner may produce a geometric path, a timed trajectory, or a sequence of controls.
Whole-body control
Whole-body control coordinates a robot’s joints and contacts to satisfy several motion and force objectives together. In humanoids, it commonly combines balance, foot motion, hand tasks, and posture subject to physical constraints.
Control barrier function
A control barrier function is a mathematical function used to express a safe set for a dynamical system and constrain control inputs so the system remains inside that set. It is commonly used as a safety filter around a nominal robot controller.
Signed distance field
A signed distance field represents a surface by assigning spatial locations a distance value whose sign distinguishes the two sides of the surface. Its zero level marks the surface, while the sign convention depends on the representation.