Online Integrity Alert Limit Determination Method for Autonomous Vehicle Navigation

Abstract

Integrity is the critical performance indicator for navigation in safetycritical applications such as autonomous vehicles. Alert limit is one of the representive parameter in integrity which defines the maximum tolerable positioning error for an operation to safely proceed. However the integrity requirements for GNSS assessment are quite different from those for autonomous vehicles. For autonomous vehicles, a reasonable alert limit needs to ensure the vehicle security and take full advantage of the space between vehicle and lane as much as possible. Based on the analysis of differences from civil aviation to autonomous vehicles, an improved alert limit determination method is proposed in this paper. The kinematic model is firstly introduced into the online determination of alert limit. The integrity risk on two sides are allocated optimally respect to the road geometry and kinematic model. The fixed cuboid bounding box is replaced by a subversive fan-shaped bounding box which is more reasonable to cover the safe-critical areas. The experiment test results compared with those of the Ford model also verified the superiority of the proposed method. Finally the paper also gives the alert limits calculated based on the Chinese standards and hopefully it could provide some references.