FastSLAM is a framework for simultaneous localization and mapping using a Rao-Blackwellized particle filter (RBPF). But, FastSLAM is known to degenerate over time due to the loss of particle diversity, mainly caused by the particle depletion problem in resampling phase. In this work, improved particle filter using geometric relation between particles is proposed to restrain particle depletion and to reduce estimation errors and error variances. It uses a KD tree (k-dimensional tree) to derive geometric relation among particles and filters particles with importance weight conditions for resampling. Compared to the original particle filter used in FastSLAM, this technique showed less estimation error with lower error standard deviation in computer simulations.

The state-of-the-art FastSLAM algorithm has been shown to cause a particle depletion problem while performing simultaneous localization and mapping for mobile robots. As a result, it always produces over-confident estimates of uncertainty as time progresses. This particle depletion problem is mainly due to the resampling process in FastSLAM, which tends to eliminate particles with low weights. Therefore, the number of particles to conduct loop-closure decreases, which makes the performance of FastSLAM degenerate. The resampling process has not been thoroughly analyzed even though it is the main reason for the particle depletion problem. In this paper, standard resampling algorithms (systematic residual and partial resampling), a rank-based resampling adopting genetic algorithms are analyzed using computer simulations. Several performance measures such as the effective sample size, the number of distinct particles, estimation errors, and complexity are used for the thorough analysis of the resampling algorithms. Moreover, a new compensation technique is proposed instead of resampling to resolve the particle depletion problem in FastSLAM. In estimation errors, the compensation technique outperformed other resampling algorithms though its run-time was longer than those of others. The most appropriate time to instigate compensation to reduce the run-time was also analyzed with the diminishing number of particles.