Guided Exploration with Proximal Policy Optimization using a Single Demonstration

Gabriele Libardi, Gianni De Fabritiis

Solving sparse reward tasks through exploration is one of the major challenges in deep reinforcement learning, especially in three-dimensional, partially-observable environments. C…