Enhanced Nearest Neighbor Method for SLAM Data Association
This paper presents an improvement to the widely-used nearest neighbor method in SLAM data association. Our approach replaces the computationally expensive Mahalanobis distance calculations between all features and measurements with Euclidean distance computations between feature estimated positions and vehicle predicted positions. This modification eliminates numerous matrix multiplication operations, resulting in a simpler algorithm implementation that significantly reduces computational complexity. The proposed method maintains comparable association performance to global nearest neighbor approaches while facilitating real-time execution of SLAM algorithms.