A NOVEL DATA ASSOCIATION TECHNIQUE TO IMPROVE RUN-TIME EFFICIENCY OF SLAM ALGORITHMS

Abstract

Simultaneous Localization and Mapping (SLAM) problem is a very popular research area in robotic applications. EKF-SLAM and FastSLAM are widely used algorithms for SLAM problem. The greatest advantage of FastSLAM over EKF-SLAM is that it reduces the quadratic complexity of EKF-SLAM. On the other hand, increasing number of estimated landmarks naturally slows down the operation of FastSLAM. In this paper, we propose a new method called as Intelligent Data Association-SLAM (IDA-SLAM) which reduces this slowing down problem. In data association step also known as likelihood estimation, IDA-SLAM skips comparing a new landmark with all of the pre-calculated landmarks. Instead of this, it compares the newly found one with only nearby landmarks that was found previously. The simulation results indicate that the proposed algorithm significantly speeds up the operation of SLAM without a loss of state estimation accuracy. Real world experiments which have been performed in two different scenarios verify the simulation results. A runtime reduction of 43% and 52% is observed respectively for each of the test environments.