Simulation of Falling Rain for Robustness Testing of Video-Based Surround Sensing Systems

A Value-Driven Design Approach for the Virtual Verification and Validation of Autonomous Vehicle Solutions

“…For this reason, the study was directed towards exploring of how assess the value of rain, snow, fog, smoke, and dirt model add-ons for Unreal Engine® and Unity 3D® engines. All these add-ons introduce 'noise' in the simulation, which must be dealt with by the virtual sensor model (see Figure 1) to ensure the correctness of the behaviour prediction (Hospach et al, 2016), such as recognizing obstacles and deciding if and how to reroute the AVS. A main reason for prioritizing this 'category' of models is that, although many simulators come with weather models, currently used image data sets are recorded largely under good weather conditions (Cordts et al, 2016).…”

A taxonomy of validation strategies to ensure the safe operation of highly automated vehicles

“…Another method to test camera sensors regarding their robustness towards rain is proposed by Hospach, Mueller, Rosenstiel, and Bringmann (2016). Their method allows to overlay a recorded video feed with simulated rain of variable strength and thus test the image processing algorithms in a SiL setup.…”

Fusing Event-based and RGB camera for Robust Object Detection in Adverse Conditions

“…Current object-detection models for autonomous driving lack the robustness to perform well in varying conditions [5]. While certain conditions have been modelled including snow [6], fog [7], rain [8], daytime and night-time transitions [5], [9], it is not possible to include all potential environmental conditions. Currently, LiDARs and RADARs are used to complement frame-based cameras to attain robustness to varying illumination and weather conditions [10], [11].…”