Method for Road Occlusions Handling in Generic Sensor Models

Abstract

Due to the increasing complexity of Advanced Safety systems, a strong endeavor is required to acquire realistic and real-time capable simulations, in order to enable robust and easily reproducible system verification in virtual environments. To make simulations reliable, high-fidelity sensor models are required. One of the approaches is to implement a generic sensor model that explicitly emulates the output of an object detection algorithm, based on high-level simulation data. However, such a model has to accurately handle object-based occlusions, to assure that shadowed objects are not detected. Various generic sensor models available in the literature already solve the problem of how to estimate occlusions, given a set of objects. Nevertheless, none of the models takes into account a road profile, i.e. hills. The method proposed in this paper provides an accurate and easy to implement road profile estimation using a set of bounding boxes. Thanks to the generated structures, a road-based shadowing can be enabled in any of the object-based generic sensor models. The obtained results clearly show the robustness and usefulness of the proposed methodology.