Autonomous Vehicle Sensor Cleaning

The automotive world is progressing fast towards autonomous vehicles making sensors one of the critical components. There is a requirement for constant exchange of information between the vehicle and its surrounding environment, which is assisted by sensors such as Camera, LiDAR and RADAR. However, exposure to harsh environmental conditions such as rain, dirt, snow and bird droppings can hamper the functioning of the sensors and in turn interrupt accurate vehicle maneuvers. Sensor-cleaning mechanisms are required to be tested under various weather conditions and vehicle operating situations. Besides wind tunnel tests, digitalizing this whole process becomes important to take decision on design changes in early vehicle development stage. This work presents a digital methodology to test the LiDAR cleaning system in the advent of mud clearing at different vehicle speeds. The cleaning mechanism consists of a telescopic nozzle placed above the LiDAR translating back and forth. As the dirt or mud is deposited on screen, water jet is sprayed from nozzle. The developed numerical method, models the motion of telescopic arm along with flat jet sprays of cleaning fluid. FVM based steady state aerodynamic simulations are employed to capture airflow while multiphase phenomenon involving interaction of air with cleaning fluid, mud particles and LiDAR screen is analyzed using an SPH solver. The results from this study and the wind tunnel show possible potential for optimization of existing cleaning systems e.g. by inspecting the surrounding airflows with increasing vehicle speeds. The qualitative comparison of these simulation outcomes indicate that there exists an inflexion point in vehicle velocity for cleaning efficiency of the system post which there is considerable degradation of its performance. Hence, the proposed numerical method provides capability of modelling distinctive designs of water cleaning systems for LiDAR which can used to evaluate and compare performance between them.