A Stochastic Physical Simulation Framework to Quantify the Effect of Rainfall on Automotive Lidar

Abstract

<div class="section abstract"><div class="htmlview paragraph">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. To this end, we combine basic probabilistic models for key rainfall parameters with Mie theory and the theory of signal detection in a Monte Carlo simulation framework. This allows to analyze and optimize a sensor’s design early in the sensor development, when physical testing is not yet possible. A challenge not addressed in this work is to include the effect of road spray water on the lidar’s performance. Combining the effect of other noise sources with the presented framework in a ray tracer is an opportunity for realistic physical lidar simulations and would allow to virtually estimate the performance of a lidar’s object detection and tracking performance. Such simulations could contribute to verify the safety of automated driving functionalities.</div></div>

Realistic LiDAR With Noise Model for Real-Time Testing of Automated Vehicles in a Virtual Environment

“…Recent works have been increasingly focusing on a more realistic modelling of the interaction between the LiDAR beams and the rain droplets, for example Berk et al takes into account a probabilistic representation of the number of drops in the volume surrounding the sensor and depending on the rainfall intensity [23]. The number of drops, ( ), is calculated by modelling the drop size distribution using an exponential distribution:…”

“…By integrating this with respect to drop size, the average number of drops in a unit volume can be determined and multiplying by the LiDAR beam volume gives the average number of drops in the beams path. This model enables the simulation of LiDAR detection performance by means of comparing probability of real object detection against probability of false alarms due to rain droplets [23].…”

“…The rain model developed builds upon the work in [23]. The drop size distribution is calculated using the exponential distribution in (4).…”

“…The drop size distribution is calculated using the exponential distribution in (4). However, this can be replaced by any distribution, should this be found to be more accurate for the application of interest [23]. This is then integrated with respect to drop size with limits of 0.5 mm and 6 mm, as raindrop size is typically in this range [36].…”

“…Equipping LiDAR to vehicles will assist towards higher levels of autonomy alongside the other sensors, such as RADAR and vision [22]. LiDAR boast excellent angular and range resolution making it an attractive addition to the vehicle sensor suite [22], [23].…”