Autonomous Pool-Playing Robot

Robotic Manipulation · 2023 · Outstanding Project

Heuristic shot planning, IK, and inverse-dynamics control sank target balls in randomized simulations.

Problem

Model a pool-playing robot well enough to plan and execute shots in randomized simulation.

Approach

Set up a Drake simulation of an IIWA arm with a cue welded to its end effector, modeling cue dynamics, ball motion, and collisions.

  • Task planning: a heuristic planner scores candidate target balls and pockets and picks the most promising shot.
  • Trajectory generation: a multi-stage end-effector trajectory lines up, strikes, and follows through, solved with inverse kinematics while accounting for ball dynamics, cue physics, and collision avoidance.
  • Control: inverse-dynamics control tracks the planned trajectory.

Results

Sank target balls consistently in randomized simulations.

Stack

DrakePython