HMPO: Human Motion Prediction in Occluded Environments for Safe Motion Planning

We present a novel approach to generate collision-free trajectories for a\nrobot operating in close proximity with a human obstacle in an occluded\nenvironment. The self-occlusions of the robot can significantly reduce the\naccuracy of human motion prediction, and we present a novel deep learning-based\nprediction algorithm. Our formulation uses CNNs and LSTMs and we augment\nhuman-action datasets with synthetically generated occlusion information for\ntraining. We also present an occlusion-aware planner that uses our motion\nprediction algorithm to compute collision-free trajectories. We highlight\nperformance of the overall approach (HMPO) in complex scenarios and observe\nupto 68% performance improvement in motion prediction accuracy, and 38%\nimprovement in terms of error distance between the ground-truth and the\npredicted human joint positions.\n