Michel Aractingi

In 2023, I obtained a Ph.D. in robotics and computer science from the National Institute of Applied Sciences (INSA) in Toulouse. I did my Ph.D. thesis at Naver Labs Europe and the Gepetto team at LAAS/CNRS. The main focus of my work revolves around learning control policies for quadruped locomotion. I am supervised by Philippe Soueres (LAAS) and Tomi Silander (NaverLabs Europe). I recieved my B.Sc from the University of Balamand in Lebanon in Electrical Engineering and my M.Sc. in data science at the Grenoble Institute of Technology INP.

During my master's thesis, I worked under the supervision of Cordelia Schmid on imitation learning for manipulation skills from visual input. After that I joined the robot navigation team in NaverLabs Europe as a research engineer working on the topic of robot navigation for indoor and crowded environments. Then, in my Phd I worked with the Solo quadruped and the MIT's MiniCheetah. I designed, implemented and transfered learned locomotion policies with reinforcement learning for both robots.

I recently started a new role as a Robotics Machine Learning Engineer at Enchanted Tools.

Email  /  CV  /  Google Scholar  /  Youtube  /  Github

I am interested in studying the role of machine learning in robotic control and perception. I prefer working with real robots and deploying learned policies in the wild. I want to use robots to push the limits of the current state of AI algorithms.

A hierarchical approach for learning locomotion policies where several aspects of the low-level behaviour can be comanded.

A deep reinforcement learning approach to learn joint-angle control for the solo12 quadruped with a state estimation network.

Learning navigation policies in densely crowded environments.

We propose to augment the model-based controller of the solo12 quadruped with a policy that modifies the gait sequence learned with deep reinforcement learning.

We study the generalization ability of visual navigation agents trained with deep RL. We propose a regularization term to improve their generalization ability.

website's template credit goes to Jon Barron