Motion Planning

Computing a sequence of poses or a trajectory from start to goal that avoids collisions and satisfies constraints.

Motion planning answers the question of how a robot moves from its current state to a goal state: finding a continuous, feasible sequence of configurations that avoids obstacles, avoids self-collision, and stays within joint limits — the classic formulation of this is the 'piano mover's problem.' Planning is usually done in configuration space, the space of every possible combination of the robot's joint values; a six-axis arm's configuration space is 6-dimensional. Common methods include grid search (such as A*), suited to low dimensions; sampling-based planning (such as RRT or PRM), which scatters random points in high-dimensional space and connects them into a path, the mainstream approach for robot arms; artificial potential fields; and trajectory optimization, which treats smoothness and time as costs to minimize. The resulting geometric path is then time-parameterized to respect velocity and acceleration limits, turning it into a trajectory that motion control tracks. End-to-end VLA models output actions directly, skipping explicit planning, but planners remain the workhorse in industrial and navigation settings.

ExampleUsing MoveIt to have a robot arm retrieve a cup from a cabinet: a sampling-based planner in OMPL finds a collision-free path in joint space around the cabinet door, which is then time-parameterized according to each joint's velocity and acceleration limits before being sent for execution.

Also called

Motion Planner

Related

Path Planning · Trajectory Planning · Configuration Space (C-Space) · Rapidly-exploring Random Tree · Collision Checking · MoveIt Motion Planning Framework

Sources

Wikipedia: Motion planning
MoveIt Docs: Motion Planning