計算コストを考慮したNDTスキャンマッチングによる位置推定の不確かさに関する研究

抄録

This paper proposes a low-cost method for estimating uncertainty in localization using NDT scan matching. The conventional uncertainty estimation involves multiple searches, which increases computation costs. To overcome this challenge, we focus our multiple searches on areas where the localization error is likely to expand. Our approach detects the possibility of the localization error increasing during the matching convergence process. We evaluated our method in an urban area with a tunnel. Compared to the conventional method, our method reduced the number of multiple search executions in the places where localization error was low. Moreover, the uncertainty of our method reflected the localization error in the tunnel as accurately as the conventional method. In future work, we will predict the possibility of localization error due to environmental changes in the scan matching.