Self-Collision Checking

Determining whether a robot's own links would collide with each other in a given pose.

Self-collision checking is a form of collision checking that only tests whether a robot's own links intersect each other at a given joint configuration — for example, the two arms of a dual-arm robot colliding, an arm hitting the torso, or the two legs touching. It is used alongside checking against environmental obstacles in motion planning, inverse-kinematics solving, and teleoperation, and can also serve as a runtime safety monitor. To save computation, an allowed-collision matrix is usually built first to rule out link pairs that never need checking: the MoveIt Setup Assistant, for instance, randomly samples 10,000 configurations by default and flags adjacent, never-colliding, or always-colliding link pairs into the SRDF. It differs from collision detection for intrinsic safety, which uses torque or momentum observers to detect a collision that has already happened; self-collision checking instead predicts geometrically, ahead of time.

ExampleFranka's FR3 Duo dual-arm setup automatically runs a self-collision-monitoring node at startup, keeping a default 0.045-meter safety margin around each link, and publishes an alert on the collision_detected topic — naming the two links involved — whenever the arms get too close.

Also called

Self-Collision

Related

Collision Checking · Collision Detection (Robot Safety) · Semantic Robot Description Format (SRDF) · Collision Geometry (Collider) · Bimanual Manipulation · MoveIt Motion Planning Framework

Sources

MoveIt Setup Assistant Tutorial: Generate Self-Collision Matrix
Franka FCI 文档：franka_selfcollision (Chinese)

As of

2026-09