Search Results

The performance of environment perceiving sensors such as e.g. lidar, radar, camera and ultrasonic sensors is safety critical for automated driving vehicles. Therefore, one has to assess the sensors’ performance to assure the automated driving system’s safety. The performance of these sensors is however to some degree sensitive towards adverse weather conditions. A challenge is to quantify the effect of adverse weather conditions on the sensor’s performance early in the development of an automated driving system. This challenge is addressed in this work for lidar sensors. The lidar equation was previously employed in this context to derive estimates of a lidar’s maximum range in different weather conditions. In this work, we present a stochastic simulation framework based on a probabilistic extension of the lidar equation, to quantify the effect of adverse rainfall conditions on a lidar’s raw detection performance.

Deep neural network models have been widely used for environment perception of intelligent vehicles. However, due to models’ innate probabilistic property, the lack of transparency, and sensitivity to data, perception results have inevitable uncertainties. To compensate for the weakness of probabilistic models, many pieces of research have been proposed to analyze and quantify such uncertainties. For safety-critical intelligent vehicles, the uncertainty analysis of data and models for environment perception is especially important. Uncertainty estimation can be a way to quantify the risk of environment perception. In this regard, it is essential to deliver a comprehensive survey. This work presents a comprehensive overview of uncertainty estimation in deep neural networks for environment perception of intelligent vehicles.