2-D Evidential Grid Mapping with Narrow Vertical Field of View Sensors Using Multiple Hypotheses and Spatial Neighborhoods

Abstract

Environment sensors with a narrow vertical field of view often fail to detect obstacles with a small vertical extent from close range. Because an inverse beam sensor model infers free space when there was no measurement, those obstacles are deleted from an occupancy grid eventhough they have been observed in past measurements. This is extremely critical if the car is driving autonomously. Our approach explicitly models these errors using multiple hypotheses in an evidential grid mapping framework that neither needs a classification nor a height of obstacles. Furthermore, we extend the grid mapping framework, that usually assumes mutually independent cells, by including information from neighboring cells. The evaluation in a freeway and a challenging scenario with a boom barrier shows that our method is superior to evidential grid mapping with an inverse beam sensor model.