Sensing Architecture for Automated/Autonomous Vehicles towards All-Condition Safety

The capabilities of a sensing system and its implementation are among the critical factors to determine the overall performance and the subsequent public acceptance of automated/autonomous vehicles. However, the uncertainty for system to work and maintain safety under undesirable conditions, such as inclement weather or low illumination, is still an obstacle to commonly accepted and legitimate applications. The safety of the automated/autonomous vehicle is in many ways determined by the performances of the subsystems. In this paper, two accuracy metrics are proposed to qualitatively assess the safety performance that each sensor can contribute. Upon the analysis of the limitations of all individual sensors, we present several novel approaches to improve the vehicle sensing architecture for achieving more reliable perception and eventually, all-condition safety.