Daniele Fontanelli

Robot learning now directly exploits the discrete geometry of polyhedral meshes, unlocking new capabilities in motion generation and planning.

Decoupling discrete task orienteering from continuous trajectory generation enables robots to maximize collaborative task yield while guaranteeing collision-free operation around dynamic human co-workers.

Robots can now localize themselves and their teammates without needing fixed infrastructure or coordinated motion, revolutionizing deployment flexibility in unstructured environments.

By intelligently directing a robot's touch, this system can more accurately and robustly identify stiff inclusions in tissue, outperforming standard Bayesian optimization techniques.