Adaptive Walking Gait for Locomotion on Terrain with Non-uniform Slope

The autonomous locomotion of a legged robot presents several challenges, such as stability and sensitivity to the ground slope. Complex sensory systems, accurate physical models, and demanding computational resources are often required to maintain balance, that together with precise servo control leads to high energy consumption levels. This paper addresses the problem of controlling the locomotion of a humanoid robot, using measurements from inertial sensors to create adaptive walking gaits. These gaits are based on the interpolation of walking gait templates, weighted by the estimated ground slope. A supervised learning method was also implemented to perform a faster recognition of the ground slope when in a static pose. This system provides a simple, yet effective approach to humanoid locomotion on sloped terrain, thus freeing computational resources for other tasks. The approach was validated in simulation, using a realistic model of a Bioloid Humanoid platform.