Recently machine learning techniques, specifically Reinforcement Learning, is being used rigorously to enable robots to perform in a desired manner. But, two challenges that exist are no safety guarantee is provided and so these techniques are still not used in safety critical systems. Secondly, most of the success shown by these learning techniques is only in simulations and the behavior in real world setting is not at all optimal/desired. The aim of this project is to borrow techniques from control theory to provide safety guarantees, smooth sim-to-real transfer, accelerated learning and robustness. Currently, we build upon the work done in Dream to Control which is a model based reinforcement learning SOTA algorithm and uses only images, rewards and actions as inputs and learns a model in a small latent space as well as actions.
