Learning to Engage: An Application of Deep Reinforcement Learning in Living Architecture Systems