Adaptive Footstep Planning for Model Predictive Control of a Quadruped Robot

Dynamic legged locomotion is a challenging control task in modern robotics, as the body is underactuated during many gaits and the contact forces of the feet are constraint due to motor limits and friction forces. A modern approach to tackle these challenges is Model Predictive Control (MPC) where the controller finds the optimal control input within a finite horizon of predicted future steps. In the literature exists a great variety of different MPC solutions. But almost all of them share, that the contact timings of the footsteps are predefined. This simplification reduces the ability to react to disturbances and leads to unnatural stepping motions for low velocities. In biology, gaits are much more variable, ensuring efficient gaits for each velocity. The objective of this work is, to implement and test an adaptive gait scheduler, which adapts the footstep timings based on the commanded velocity and the current state of the robot, to react to external disturbances and reduce the required cost of transport.