A Formal Experiment to Assess Pedestrian Detection and Tracking Technology for Unmanned Ground Systems

An important area of investigation in robotics perception and intelligent control concerns the ability to detect, track, and avoid humans operating in proximity to an unmanned ground vehicle (UGV). Under the Army Research Laboratory (ARL) Robotics Collaborative Technology Alliance (RCTA), ARL and other member organizations have developed algorithms focused on human detection and tracking, which leverage program advances in stereovision and LADAR. A recent assessment conducted by ARL and the National Institute of Standards and Technology (NIST) exercised these technologies under relevant conditions. This paper highlights technology advances demonstrated in this investigation. The most significant findings are that pedestrians can be reliably detected and tracked and that with the inclusion of temporal filtering on algorithm reports, incidences of misclassification of other objects as pedestrians can be dramatically reduced.