Hajid Alsubaie

Biography

Hajid Alsubaie's research interests include model reduction for deterministic, stochastic, and nonlinear dynamical systems. 

His theoretical work focuses on model-free, decentralized multi-agent reinforcement learning in discounted Markov potential games, where agents learn from local bandit feedback without a common prior or explicit communication. Value-function learning and policy updates run on two timescales via asynchronous actor-critic methods. The aim is large-scale asymptotic and convergence analysis under weaker assumptions, with robustness and safety cast as mathematical constraints that yield provable guarantees.