Continual Model-based Reinforcement Learning – TAKEN

Model-based reinforcement learning (MBRL) is emerging as a powerful paradigm for building intelligent agents capable of operating in the physical world. Most existing MBRL algorithms assume that the environment is stationary, i.e., that its behaviour doesn’t change so that the learnt world model and the strategies than depend on it remain optimal over time. But what happens if the environment changes (continuously) after the model is built? This project will explore how continual learning strategies might be used to update a world model as the environment changes. The project will undertake a comparative analysis of existing proposals to support continual learning and assess their applicability in a context where the system is interacting with/controlling the physical world, e.g., an automated vehicle.