Robot Simultaneous Localization and Mapping Based on Non-Linear Interacting Multiple Model

Abstract

To investigate robot simultaneous localization and mapping (SLAM) in the unknown environment, the non-linear interacting multiple model (IMM) SLAM algorithm is applied to solve the problem concerning the statistical property mutation of a system. The key point of this algorithm is to use non-linear Gaussian model to approximate non-linear and non-Gaussian model so that robot Simultaneous Localization and Mapping can be achieved. Each model employs the extended Kalman filter (EKF) algorithm to linearize the non-linear system and uses the non-linear interacting multiple model algorithm in each step to get fusion estimated value. The Monte Carlo simulation results indicate that when the process covariance and observation covariance change, the non-linear interacting multiple model SLAM algorithm has better estimate precision compared with EKF-SLAM algorithm and Fast-SLAM algorithm.