Welcome to the Poonawala Lab website. Our lab works at the intersection of control theory, machine learning, artificial intelligence and robotics. Our over-arching interest is in autonomous robot navigation and manipulation.
Robotics and control experts have designed several algorithms to solve problems in this area (deep reinforcement learning, feedback linearization, model predictive control), however they rely on human validation in some form or the other before deployment. In deep reinforcement learning, humans have to develop a sensible reward function and validate the trained object, largely due to issues of reward hacking. In more control theoretic approaches, the human has to specify a Lyapunov function that proves system properties, or a control Lyapunov function that will yield the right control.
Our lab aims to develop automated algorithms that robots use to design provably safe controllers for unseen environments using data, without human intervention. Current algorithms that provide all three features (provably safe, data-driven, human-oversight-free) in their controllers are highly application specific. We believe that expertise in machine learning, dynamics, control theory, and artificial intelligence is critical to developing more general algorithms.
Research OpportunitiesOur lab is growing and we invite those with interest in robotics research to join us. We seek individuals with knowledge and interest in
- Control of Mechanical Systems
- Machine Learning for Perception and Control
- Hybrid Control Systems
- Correct-by-Construction Control using Formal Methods and Model Checking
- Manipulation and Grasping with Robot Arms
- Planning under Uncertainty: Markov Decision Processes, Partially Observable MDPs, RRT*, etc
- a one-page resume,
- up to three papers or technical reports primarily written by you, and
- a transcript.
Preference will be given to candidates with strong records in research/coursework related to mathematics, optimization, control, or robotics. Individuals with significant theoretical or experimental experience are encouraged to apply.