In this effort, Methods for online real-time learning that are robust to modeling errors and abrupt changes in the dynamic models are developed via integration of model validation, model-free and model-based reinforcement learning techniques in a model-aware reinforcement learning framework. The intellectual merit of the proposed work stems from the emphasis on real-time estimation, synthesis and execution. The focus on real-time operations and simultaneous learning and execution requires integration of RL with control theoretic considerations such as system stability during learning phase, that are rarely studied in the machine learning literature.


Get in touch
(405) 744-5900