Modeling Methodology and In-field Measurement Setup to Develop Empiric Weather Models for Solid-State LiDAR Sensors

Zusammenfassung

With the automotive industry’s dedicated roadmap towards partly automated driving, the responsibility for reliable environmental perception moves from the driver to the vehicle’s advanced driving assistance systems (ADAS). However, with steadily growing system complexity, the required test mileage to certify new driving functions increases to an unworkably high level. In order to validate ADAS functions like lane change assist (LCA), automated emergency breaking (AEB), or even path planning virtually, there is a strong demand for high fidelity sensor models which are capable of simulating automotive Radar, LiDAR as well as camera sensor perception in real time while providing realistic, artificial sensor raw data. Yet, especially LiDAR models mostly lack the capability of replicating the impact of specific weather characteristics, although optical sensors in particular are heavily influenced by precipitation, fog and sun irradiance. Furthermore, there is - in contrast to numerous publicly available LiDAR datasets in differing driving situations - a strong lack of datasets which are annotated with quantitative weather data such as particle size and velocity distribution in order to develop and validate such models. Hence, within this work, an automated infrastructure setup for targeted measurement of time-correlated LiDAR and weather data is presented with the aim to develop and calibrate weather models, which can eventually be used to augment virtual LiDAR data from raytracing capable driving simulation suits as well as real data, recorded under ideal weather conditions. In addition to that, the considerable effect of varying precipitation rates on an automotive Flash LiDAR system was demonstrated based on first measurements and quantified by calculating the pixel-wise temporal coefficient of variation for measured depth and intensity, reaching up to approximately 50% and 350%, respectively.

Publikation
2022 IEEE 2nd International Conference on Digital Twins and Parallel Intelligence (DTPI)