Implementation of a Deep Reinforcement Learning algorithm, Proximal Policy Optimization (SOTA), on a continuous action space openai gym (Box2D/Car Racing v0)