16 January. 2021
Legged robots and in particular quadruped ones, are an emerging, promising technology for supporting humans in settings such as disaster response, agriculture, decommissioning, and inspection. During this internship, we will conduct research on AI for such robots. Examples of projects include: 1) intelligent trial-and-error (Cully et al., 2015) for damage recovery; 2) learning agile locomotion with deep reinforcement learning; 3) automated curriculum generation for creating robust skills; 4) perception-driven gait switching; 5) meta-learning for fast adaptation in new environments; 6) model-predictive control; 7) sim2real approaches; 8) lifelong simultaneous localization and mapping; 9) motion retargeting.
Cully, A., Clune, J., Tarapore, D., & Mouret, J.-B. (2015). Robots that can adapt like animals. Nature, 521(7553), 503-507.
Good programming skills (C++ or Python). Machine Learning (e.g., reinforcement learning, deep learning). Robotics ( simulators [e.g., Gazebo, PyBullet], ROS, control theory). Knowledge of Quality-Diversity algorithms (e.g., MAP-Elites) or physically-based animation will be considered an advantage.