Integrated Walk and Kick for a Humanoid Robot Using Model Predictive Control

• Gabriel P. L. Ferreira ITA
• Marcos R. O. A. Maximo ITA

Resumo

In humanoid robot soccer, the control of the kick motion plays an important role in various situations from ball challenges and passes to shots on goal. Improving kick performance requires not only increasing its power, but also reducing its duration and enhancing accuracy, all while maintaining stability. For instance, methods that require the robot to completely stop before kicking are disadvantageous in ball challenge scenarios. Additionally, approaches that rely on predefined swing foot trajectories demand highly precise positioning relative to the ball, which increases execution time and may compromise kick accuracy. This paper contributes by proposing methods for generating both the swing foot trajectory and the robot's center of mass trajectory to enable accurate and stable kicking during continuous walking. The swing foot trajectory is generated online as a B-spline, based on time-varying inputs such as the ball position and the desired shot direction. Meanwhile, the torso trajectory is generated by a model predictive controller as a continuation of the walking motion. The method was validated in simulation and real-robot experiments, demonstrating successful kick execution without interrupting gait while preserving kick power.

Palavras-chave: Legged locomotion, Torso, Accuracy, Humanoid robots, Stability analysis, Real-time systems, Trajectory, Splines (mathematics), Foot, Sports, Humanoid Walking, Kicking, Humanoid Robot, Robot Soccer