Abstract

In this talk, I will present my research on enhancing robotic motion and control across diverse and challenging scenarios. First, I will explore the generation of optimal jumping motions using Reinforcement Learning (RL), integrating physical insights to guide the learning process. This significantly reduces the training time compared to standard end-to-end approaches. Second, I will illustrate the development of novel robotic platforms that combine ropes and legged systems to address the challenges of maintenance in remote, hazardous environments. By leveraging optimal control techniques, I designed a computationally efficient planning algorithm for precise and adaptive jumping motions. Finally, I will investigate tracked vehicles for precision agricultural robotics, focusing on slippage-aware planning and a shared control strategy that ensures seamless transitions between human and automated inputs. These advancements contribute to safer, more efficient and adaptable robotic solutions for critical real-world application, from infrastructure maintenance to sustainable agriculture.

About the Speaker

Michele Focchi is a world recognized expert in motion planning and control of quadruped robots, with 16 years of experience in the field of robotics. He is particularly known for his pioneering work on heuristic locomotion in unstructured terrains. Currently, he holds the position of scientific advisor for ALL3. He received both his Bsc/Msc in Control System Engineering from Politecnico di Milano. In 2013 he got a PhD in robotics where at  Istituto Italiano di Tecnologia where he was co-founder of the Dynamic Legged System (DLS) lab, an international research team dedicated to the development of quadruped robots and the study of their locomotion.
His research interests lie  at the intersection between control, optimization and machine learning with a focus on enhancing the performance of quadruped robots in challenging environments, by using optimization-based techniques. He explored innovative robotic platforms, such as a rappelling robot, to address specific challenges like hydro-geological risk reduction. More recently, he has also begun investigating control strategies for tracked robots in agricultural applications. He organized several workshops on scientific topics and  has given more than 20 invited presentations at international workshops, and dissemination events. In terms of technology transfer, in 2015 he co-founded the MOOG-IIT joint lab to develop advanced software and control solutions for autonomous robots. He played important roles in several high-profile industrial and academic projects, such as ECHORD++, INAIL and, more recently, ANT with the European Space Agency. He is author or co-author of more than 53 scientific papers on international journals and conferences, with high citation records  and supervised numerous master’s and PhD theses. From 2022 to 2024, he was at the University of Trento, where he taught introductory courses on robotics. Additionally, he serves as an Associate Editor for the RA-L journal and the ICRA conference.