Solving the Bipedal-Walker Gym environment.

As shown in the image below, the algorithm achieves an average of 300+ rewards in 100 episodes, the required criteria for "solving" the environment.

As shown in the video, the agent can successfully walk to the end of the terrain without falling, on every single run.

bipedal_walker_demo.mov