2014-169 – Pedestrian motion prediction from laser-based tracking data for driving applications

The goal of this invention is to be able to predict, or forecast, the motion of pedestrians in city traffic conditions. The technology used existing prediction models to convert the partially observed trajectory of a pedestrian to a probability map of possible locations of the pedestrian in the future (see figure). The partial trajectory is obtained used existing technology for tracking pedestrian data from a laser range finder such as a SICK scanner. The output of the system is a probability map that is recomputed at regular intervals, fast enough for the predictions to be generated as the car moves at normal speed for busy city traffic. To summarize, the invention contributes to a system that includes:- A tracker, in this case tracking using a laser ranger finder- An algorithm to generated probability maps of future motion for each individually tracked pedestrian- Synchronization between tracking and prediction to allow for constant re-evaluation of the prediction maps as the vehicle moves. Scott McEvoy smcevoy@andrew.cmu.edu 412-268-6053