Human Motion Prediction Based on Object Interactions

Publication Authors L. Bruckschen; N. Dengler; M. Bennewitz
Published in Proceedings of the European Conference on Mobile Robots (ECMR)
Year of Publication 2019
Abstract

In this paper, we consider the problem of predicting the navigation goal of a moving human in an indoor environment. Knowledge about this goal can greatly increase the efficiency of robots acting in the same environment as interferences can be avoided and assistance quickly provided if necessary. Often the navigation goal depends on the previous action of the human and the object the human has interacted with before. Thus, the information about previous object interactions can be used to infer possible objects the human will interact with next, which in term can be used to predict the current navigation goal. We propose to learn a probability distribution of subsequent object interactions and present a framework that utilizes the learned transition model as well as observations of the human’s location and pose for the prediction of their movement goal. As we show in various experiments, the information about transition probabilities of object interactions leads to reliable predictions of the navigation goal and improves the accuracy compared to prediction approaches that rely only on spatial information and do not consider object interactions. Furthermore, we demonstrate how the prediction can be used to realize foresighted robot navigation.

Type of Publication Conference Proceeding
Lead Image No image
Lead Image Caption
Text
Images
Teaser Image 1
Teaser Image 2 No image
Files and Media
Local Video File
Settings
Versioning enabled yes
Short name human-motion-prediction-based-on-object-interactions
Layout
Blocks { "e238dda9-b399-4262-b67b-34ac6faab0d6": { "@type": "slate" } }
Blocks Layout { "items": [ "e238dda9-b399-4262-b67b-34ac6faab0d6" ] }
Options
Categorization
Related Items
Contents

There are currently no items in this folder.