Socially Aware Motion Planning with Deep Reinforcement Learning

摘要: For robotic vehicles to navigate safely and efficiently in pedestrian-rich environments, it is important model subtle human behaviors navigation rules (e.g., passing on the right). However, while instinctive humans, socially compliant still difficult quantify due stochasticity people's behaviors. Existing works are mostly focused using feature-matching techniques describe imitate paths, but often do not generalize well since feature values can vary from person person, even run run. This work notes that challenging directly specify details of what (precise mechanisms navigation), straightforward (violations social norms). Specifically, deep reinforcement learning, this develops a time-efficient policy respects common norms. The proposed method shown enable fully autonomous vehicle moving at walking speed an environment with many pedestrians.