Improvements in Pedestrian Movement Prediction by Considering Multiple Intentions in a Multi-Hypotheses Filter

Florian Particke, Markus Hiller, Christian Feist, Jörn Thielecke

Abstract:	Fully automated vehicles and mobile robots have to operate in a shared environment with pedestrians in the future. To minimize the risk for pedestrians, it is very important to track them in a precise way. One important information source is the intention of the pedestrian. For the integration of the intention information with conventional tracking algorithms, a Multi-Hypotheses filter is proposed, using a generalized potential field, which can be modeled using pedestrian movements. As the intention of the person is unknown, different hypotheses for the intention of the pedestrian are considered. The MultiHypotheses filter is used for the prediction of pedestrian trajectories and compared with a simple Kalman filter. The evaluation is performed in dependence on the free parameters, which are the prediction time, the tracking time and the measurement quality (modeled by additional additive white Gaussian noise). The proposed approach is evaluated using real camera data from a simple scenario in the Edinburgh Informatics Forum. For evaluation, the root mean square error and a confidence score, which is based on the normalized entropy, are considered. The Multi-Hypotheses based prediction outperforms the simple Kalman filter over the whole range of measurement quality, prediction and tracking time horizons in the case of the root mean square error and in the case of the confidence score.
Published in:	2018 IEEE/ION Position, Location and Navigation Symposium (PLANS) April 23 - 26, 2018 Hyatt Regency Hotel Monterey, CA
Pages:	209 - 215
Cite this article:	Particke, Florian, Hiller, Markus, Feist, Christian, Thielecke, Jörn, "Improvements in Pedestrian Movement Prediction by Considering Multiple Intentions in a Multi-Hypotheses Filter," 2018 IEEE/ION Position, Location and Navigation Symposium (PLANS), Monterey, CA, April 2018, pp. 209-215.
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In