Fault Tolerance from Formal Analysis of a Data Fusion Mechanism

Multi-sensor perception systems are starting to be used incritical applications, such as in drones and ADAS (AdvancedDriver Assistance Systems). However, complete validation ofcomplex perception systems is difficult to achieve. In this paperwe examine these systems through the lens of an alternativedependability method, namely fault tolerance. We apply a formalanalysis on a belief function data fusion mechanism to providefault tolerance. By analyzing certain parameters related to thedata fusion process, we show that it is possible to offer faulttolerance services suitable for multi-sensor perception systems, including detection, recovery, and fault masking.