Title: Perceptive Locomotion through Full-dynamics MPC and Mixed-integer optimisation.
Abstract: Dynamic quadrupedal locomotion in a complex environment requires efficient environmental perception, precise footstep planning, collision avoidance, and a robust locomotion controller. We propose a complete pipeline from detection to motion generation in clustered environment based on many different works. This work has been carried out for the European MEMMO project. This synthesis aims to find a trade-off between different methods to generate complex motion while staying relatively simple and computationally efficient for real-time application. As usually done in perceptive locomotion, we decompose the problem into smaller pieces to face computational challenges. Convex planes are first extracted from the on-board depth camera. A mixed-integer optimisation handles the combinatorics problem related to the choice of surfaces. Following this first layer, the foot location is adapted locally inside the surface and a collision-free trajectory is then optimized for each foot. Finally, a model predictive controller (MPC) based on the full dynamics of the robot generates the whole-body motion. We validate our approach on various scenarios including stairs, gaps, holes, stepping stones.