Glossary · Robotics and control
Motion planner
Also called planner, footstep planner, step planner
The part of a system that decides which movement to make next to reach a goal, leaving the details of making it to a controller.
Reaching a goal usually splits into two questions: what to do next, and how to do it. A motion planner answers the first. It looks at the goal and the current state and chooses the next movement: a step in some direction, a turn, a path around an obstacle. A controller, or a learnt policy, answers the second: it turns the chosen movement into motor commands that work with real physics.
Walking robots often use a footstep planner: it picks where each foot should land, and a balance controller makes the body get there. Video games do the same with animated characters: the game logic chooses “turn left, then walk forward”, and the animation system plays the matching clips.
A planner can be a search over many possible paths, or a few simple rules. Rules are enough when the world is an empty floor.
In this project
The fighter’s walk uses a rule-based planner. At the start of each beat of the fighter’s rhythm it looks at the target, a spot for the lead foot and a direction to face. If the facing is more than 30° off, it plans a pivot, turning the way that needs fewer pivots. Otherwise it plans one step straight towards the spot, in any of 16 directions, sized to land on it, with up to 10° of steering. The spot counts as reached within 15 cm, or after four beats within 25 cm, so the fighter doesn’t circle round it. The policy then makes the planned step happen on the simulated body (see Footwork).