Human-Aware Robot Trajectory Planning with Hybrid Candidate Generation: Leveraging a Pedestrian Motion Model for Diverse Trajectories

Robots navigating in populated dynamic environments require dedicated motion planning algorithms. We propose a human-aware local trajectory planner that alleviates the discomfort of humans surrounding the robot. Our method relies on a hybrid approach to trajectory generation and employs cost functions evaluating robot navigation task performance, robot motion naturalness, and humans' perceived safety. We compared our method with state-of-the-art classical and human-aware trajectory planners using the TIAGo robot in simulated and real-world environments. The algorithms tested were quantitatively assessed for both navigation performance and human discomfort using the Social Robot Planner Benchmark. Our approach was implemented as an open source plugin for the local trajectory planner of the Robot Operating System's navigation.