Count-Based Exploration via Embedded State Space for Deep Reinforcement Learning

Figure 3

Illustrations of the Rllab task used in continuous control experiments, namely, CartPoleSwingup, HalfCheetah, and SwimmerGather.