A SLAM Algorithm for Intelligent Vehicle based on the Square Root Central Difference Particle Filter and Per-Particle ICNN-ML Data Association

摘要

Intelligent vehicle plays an important role in intelligent transportation systems. Developing a high-performance SLAM algorithm can provide the key technology for intelligent vehicle achieving autonomous navigation in an unknown environment. In simple terms, SLAM problem includes two issues of state estimation and data association. Hence, a SLAM algorithm based on the square root central difference particle filter (SRCDPF) and per-particle ICNN-ML (individual compatibility nearest neighbor-maximum likelihood) data association method is proposed in this paper to solve the SLAM problem. In the proposed algorithm, the per-particle ICNN-ML data association method provides the key information for the state estimation. The state of each feature in the map is estimated by using central difference Kalman filter (CDKF). The square root central difference particle filter (SRCDPF) is used to compute the mean and covariance of the vehicle state. The simulations result in a large-scale environment which is designed to emulate a real campus scene and the experimental result with standard dataset have verified the feasibility and effectiveness of the proposed SLAM algorithm.