Jean-Baptiste Bouvier

Postdoc @UCBerkeley Mechanical Engineering. Robotics, Safe Control, Reinforcement Learning, Machine Learning

[Jean-Baptiste Bouvier @JB_Bouvier__]

If you work on reliable robotics, we want to hear your take on how GenAI impacts the safety and security of robots.
We are organizing a workshop on June 21st at #RSS2025 and we are awaiting your short paper submissions!

Deadline: May 20th
Website: https://t.co/vC8PrwEadl

[Jean-Baptiste Bouvier @JB_Bouvier__]

@gokulp01 Thanks Gokul!

[Jean-Baptiste Bouvier @JB_Bouvier__]

Very excited to introduce my latest work on generating feasible robot trajectories with diffusion models!

Website: https://t.co/qvIg6RyfmB
Paper: https://t.co/c2QSPxnWZO

Thread ↓

[Jean-Baptiste Bouvier @JB_Bouvier__]

We also showed that our approach generates better open-loop robot trajectories in hardware 🤖

[Jean-Baptiste Bouvier @JB_Bouvier__]

We deployed our approach on several MuJoCo simulations where our diffusion transformer generates long-horizon dynamically feasible trajectories in open-loop!

[Jean-Baptiste Bouvier @JB_Bouvier__]

I am also excited to meet and talk with other researchers at the Faculty Candidate Poster session on Monday!

[Jean-Baptiste Bouvier @JB_Bouvier__]

Next week I will be attending CDC in Milan to present our work on enforcing hard constraints of high relative degree in reinforcement learning. We apply our POLICEd RL approach to land the Space Shuttle!
Website: https://t.co/zLYMcsZmOy

[Jean-Baptiste Bouvier @JB_Bouvier__]

I'm excited to share that I will be presenting our work with Prof.@negar_mehr and Kartik Nagpal at #RSS2024!
Our work POLICEd RL enables to learn a robot policy provably satisfying a hard constraint in closed-loop with a black-box environment.

Website: https://t.co/ecyhXCdnkT