Hardware reliability in robo-taxi

Self-driving technology is reshaping the future of our transportation. It's full of opportunities and challenges. In this paper, we will share our view of hardware reliability in autonomous vehicles, especially robo-taxi. Compared with conventional vehicles, robo-taxi has different use cases, such as longer daily operation hours and mileage. Existing industry standards might not work well for the qualification of robo-taxi. Based on our operational design domain, we customized accelerated stress profiles by using road instrumentation for vibration and leveraging weather station data for temperature cycle, temperature humidity and UV exposure. This kind of stress profile customization can balance the time to market and design effort and avoid over-design or under-design. Vibration testing of network switch and shock testing of boards for multimedia and networking applications were used as case studies to show the impact of stress profiles. Robo-taxi introduces many new technologies for sensing and computing (e.g. camera, high performance computer, lidar, radar, and thermal cooling system). Additionally, to save cost and reduce development time, some off-the-shelf and non-auto grade components were incorporated. Due to these, new stresses were introduced for qualification and new failure modes have been observed. Air bubble in coolant, camera connector sealing material down selection, cold plate buckling, coolant leakage, LED failure, Lidar bracket crack, mechanical relay failure, paint blistering, PCB corrosion failure, power supplier unit failure, and Radar thermal step failure were presented. Analytical modeling, FEA modeling, and experimental characterization techniques like FTIR, I-V curve, SEM/EDX, 3D xray, and 3D optical scanner were combined to understand the root causes. Robo-taxi has an expensive BOM and fast development pace. Because of these, less sample size and shorter test time are required. Several methodologies were presented to handle this requirement.