Learning To Walk Via Deep Reinforcement Learning