Home

The candidate will work on Tasks 2-3 of the project, addressing the problem of model-based reinforcement learning in non-stationary environments using deep learned models. The key objectives are: 1. Conduct an extensive survey of state-of-the-art on model-based reinforcement learning in non-stationary environments, where non stationarity occurs from the presence of other agents in the environment, and planning using the learned models. 2. Develop new methods for learning disentangled latent representations for such processes that can be used to enable the emergence of communication. 3. Extend MCTS planning algorithms to handle the models learned in 2.