Maulik Bhatt

I am a Ph.D. student in the Department of Mechanical Engineering at UC Berkeley, where I am advised by Prof Negar Mehr. I am a part of the ICON lab. I started my Ph.D. at UIUC as a part of the Coordinated Science Laboratory. Later, I moved to UC Berkeley with my advisor, Prof Negar Mehr, starting in January 2024.

In Summer 2025, I interned at Amazon Robotics, where I worked on learning based motion planning algorithms for fully autonomous warehouse robot, Proteus.

Previously, I graduated from IIT Bombay with an interdisciplinary dual degree(Bachelor's in Aerospace Engineering + Master's in Systems and Control Engineering). During my stay at IIT Bombay, I was awarded Panasonic Scholarship, Undergraduate Research Award, and Technical Hostel's Commendation.

Email  /  CV  /  Google Scholar   /   Twitter   /   Github

I am interested in developing autonomous robots that can work alongside humans and other robots. My research focuses on enabling agents to reason about other agents in the environment and effectively coordinate their motion. To this end, I have worked on approaches grounded in diffusion models and game theory, developing scalable algorithms that achieve real-time performance and robust multi-agent coordination.

MIMIC-D is a framework for multi-agent coordination using only local observations that effectively captures multi-modality in the expert data.

Our algorithm successfully estimates observation space of the agents and is able to imitate their behavior.

Our proposed method reduces the computation time of multi-modal equilibria upto 50%.

We introduce the notion of noisy Nash equilibrium and provide forward solutions and inverse solutions to estimate the objectives of the agents from demonstrations.

Our constrained trajectory planner is 20 times faster than the state-of-the-art

Our algorithm outperfroms the state-of-the-art.

To the best of our knowledge, this is the first pose estimator that is proven to be almost globally asymptotically stable in discrete-time

To the best of our knowledge, this is the first attitude estimator that is proven to be almost globally asymptotically stable in discrete-time and considers multi-rate measurements

Developed an algorithm to explore an unknown region while avoiding obstacles

Implemented a Dynamic Image-Based Visual Servo Control technique for safe autonomous landing of drone on moving ground vehicle

Climbing, Hiking, Cooking, Flute