Hitter: Humanoid Table Tennis Robot

Technical Analysis Humanoid robots playing table tennis thanks to HITTER Can humanoid robots play table tennis agains humans and act like humans? Here comes "HITTER: Humanoid Table Tennis Robot", a hierarchical framework for humanoid table tennis that integrates a model-based planner for ball trajectory prediction and racket target planning with a reinforcement learning-based whole-body controller. Motivation of the Work Humanoid robots have recently demonstrated impressive capabilities in basic movements like locomotion (walking) and motion imitation. However, most current research still focuses on control in free space or scenarios involving only static interaction with the ground, such as maintaining balance or mimicking human poses. A significant challenge is rapid interaction with fast-moving objects in dynamic environments through manipulation. Tasks like table tennis require perception, prediction, planning, and action to occur within sub-second reaction times, demanding coordinated control across many joints. While specialized robotic arms have achieved progress in robotic table tennis, general-purpose humanoid robots previously demonstrated were often constrained to stand still without agile locomotion. This limitation severely restricted their effective hitting range across the table. The researchers were motivated to use table tennis as a unique and stringent testbed for robotics due to three critical demands: High Dynamism: It requires fast ball trajectory…