Safety analysis of Autonomous Ground Vehicle optical systems: Bayesian belief networks approach

Autonomous Ground Vehicles (AGV) require diverse sensor systems to support the navigation and sense-and-avoid tasks. Two of these systems are discussed in the paper: dual camera-based computer vision (CV) and laser-based detection and ranging (LIDAR). Reliable operation of these optical systems is critical to safety since potential faults or failures could result in mishaps leading to loss of life and property. The paper identifies basic hazards and, using fault tree analysis, the causes and effects of these hazards as related to LIDAR and CV systems. A Bayesian Belief Network approach (BN) supported by automated tool is subsequently used to obtain quantitative probabilistic estimation of system safety.