ICM: An efficient data association for SLAM in stochastic mapping

In this paper, iterative classification matching (ICM), a novel practical data association method, is proposed. ICM is an iterative approach to solve the data association problem which reconsiders the established observation-feature pairing and applies the quaternion approach to yield a least squares matching vector. The map features which are not associated with any observations are updated then by the obtained least squares matching vector to weaken the influence of the inaccurate vehicle pose estimation. Finally, the updated feature set and the unassociated observations are taken as a group of new inputs to perform the iteration again. The iteration is terminated until the discrepancy in mean square error falls below a preset threshold specifying the desired precision of the matching. Results of simulation experiments show that the proposed ICM method is an efficient solution to data association. Unlike ICNN (individual compatibility nearest neighbor), ICM can provide a robust solution in both simulated and real outdoor environments. Simultaneously, the computational cost of the proposed ICM algorithm is much lower than JCBB (joint compatibility branch and bound).