Andrea Baisero

I am a PhD student at the Lab for Learning and Planning in Robotics (LLPR) at Northeastern University, where I work on partially observable reinforcement learning.

Currently seeking positions in reinforcement learning research, with particular interest in roles involving partial observability and multi-agent systems.

Research Focus

I specialize in model-free reinforcement learning for partially observable tasks, where agents must develop sophisticated information-gathering strategies and selective memory mechanisms. My research focuses on privileged information frameworks that exploit additional information during training that is unavailable during execution, (e.g., simulated training, training in controlled environments, centralized training with decentralized execution (CTDE) for multi-agent systems).

During my PhD, I worked extensively on privileged information approaches, identifying critical limitations in prior work and proposing novel solutions that achieve superior performance with more efficient models. My contributions span theoretical advances in understanding privileged information use and practical algorithms that bridge the gap between training and deployment constraints.

I believe partially observable reinforcement learning represents an underexplored frontier with immense practical importance. My research combines rigorous theoretical analysis with empirical validation in challenging domains, contributing to both fundamental understanding of learning under uncertainty and practical solutions for real-world applications where complete observability is unrealistic.

Research Interests: Partial observability, multi-agent control, privileged training frameworks, model-free and model-based RL, combining RL with planning solutions, information-gathering, and causality.