Carlos Mastalli
Bio: Carlos Mastalli is an Assistant Professor at Heriot-Watt University, Edinburgh, U.K. He is the Head of the Robot Motor Intelligence (RoMI) Lab affiliated with the National Robotarium and Edinburgh Centre for Robotics. Carlos is also a Research Scientist at IHMC, USA. Previously, he conducted cutting-edge research in several world-leading labs: Istituto Italiano di Tecnologia (Italy), LAAS-CNRS (France), ETH Zürich (Switzerland), and the University of Edinburgh (UK). His research focuses on building athletic intelligence for robots with legs and arms. Carlos’ research work is at the intersection of model predictive control, numerical optimization, machine learning, and robot co-design.
Abstract
Pushing the robot limits when synthesising legged locomotion maneuvers requires generating actuation-aware whole-body motions and footsteps ahead of a perceived environment. However, there are significant computation limitations when doing so. First, the combinatorial nature of selecting contact regions in a perceptive locomotion pipeline is high. Second, whole-body motion generation poses a large and nonlinear optimization problem. To leverage these limitations, we advance both mixed-integer and nonlinear optimal control formulations and integrate them within a novel perceptive locomotion pipeline. We validate our perceptive locomotion pipeline in a wide range of terrain conditions that push the limits of the ANYmal B robot.