Abstract
In this paper the influence of different information on the performance of pedestrian motion models is evaluated. For this, besides of dynamic and physiological constraints, additional information as head orientation, typical pedestrian motion behavior and environment informations are used. A vehicle sensor system, consisting of a 12Mpx camera and Lidar provides the pedestrian and environment information in the considered test cases. Based on this, four pedestrian models, each based on different information, are used to predict the future position of the pedestrian. The predicted area is presented by a probability grid. The results show the performance of the models for evaluation parameters as prediction error, standard deviation and collision probability. Earlier works show, that the models have different performance depending on the prediction time. So a new introduced system fuses the probability map of each model to achieve a continuous prediction performance.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call