Prediction of Human Full-Body Movements with Motion Optimization and\n Recurrent Neural Networks

2019/10/04 by Philipp Kratzer, Kratzer, Philipp, Marc Toussaint +3 · 2 citations
Computer Science · Engineering · #Human Pose and Action Recognition #Human Motion and Animation #Anomaly Detection Techniques and Applications

paper · pdf · doi:10.48550/arxiv.1910.01843

Abstract

Human movement prediction is difficult as humans naturally exhibit complex\nbehaviors that can change drastically from one environment to the next. In\norder to alleviate this issue, we propose a prediction framework that decouples\nshort-term prediction, linked to internal body dynamics, and long-term\nprediction, linked to the environment and task constraints. In this work we\ninvestigate encoding short-term dynamics in a recurrent neural network, while\nwe account for environmental constraints, such as obstacle avoidance, using\ngradient-based trajectory optimization. Experiments on real motion data\ndemonstrate that our framework improves the prediction with respect to\nstate-of-the-art motion prediction methods, as it accounts to beforehand unseen\nenvironmental structures. Moreover we demonstrate on an example, how this\nframework can be used to plan robot trajectories that are optimized to\ncoordinate with a human partner.\n

Cited by

• On Using Neural Networks to Learn Safety Speed Reduction in Human-Robot Collaboration: A Comparative Analysis
• Learning-Based Safety-Aware Task Scheduling for Efficient Human-Robot Collaboration

Related

• Planning Coordinated Human-Robot Motions with Neural Network Full-Body Prediction Models
• Human Motion Prediction using Semi-adaptable Neural Networks
• On human motion prediction using recurrent neural networks
• Context-aware Human Motion Prediction
• A Data-Efficient Approach for Long-Term Human Motion Prediction Using Maps of Dynamics