Degradation state detection and local map optimization for enhancing the SOTIF of map-matching-based fusion localization system

Abstract

In scenarios involving local map missing and local environmental changes, map-matching localization systems (MMLS) may experience degradation in localization due to the deficiency of texture and geometrical features. This consequently leads to a decrease in localization accuracy or divergence in a map-matching-based fusion localization system (MMBFLS). To address this issue and ensure the safety of the intended functionality of autonomous vehicles, this study proposes a degradation state detection and local map optimization-based fusion localization frame. To detect the degradation state of pose estimation in MMLS, the analysis of the mathematical mechanisms underlying the localization degradation in MMLS is conducted first. Subsequently, a degradation state detection method based on incremental matrix eigenvalue analysis is proposed, which leverages the magnitude of individual eigenvalues of the incremental matrix to determine the corresponding degradation state and isolates it accordingly. Furthermore, we employ sliding window-based factor graph optimization to maintain a local map that matches Light detection and ranging point clouds, compensating for the degraded states and thereby ensuring the localization performance of MMBFLS. Real-world sensor data is collected to validate the proposed method. The results indicate that the accuracy of the proposed degradation detection method reached as high as 94.67%. Moreover, the proposed localization framework can effectively ensure the localization accuracy of MMBFLS in typical highway scenarios.