A UPF-PS SLAM Algorithm for Indoor Mobile Robot With NonGaussian Detection Model

Abstract

In this article, we propose a simultaneous localization and mapping (SLAM) framework that combines the methods of the unscented particle filter (UPF) and particle smoothers (PS). The UPF is used to estimate the position and orientation of the mobile robot and PS are used to update the locations of landmarks. We show that UPF is capable of estimating the pose of the robot more consistently and accurately than generic particle filter and unscented Kalman filter (UKF) do, especially when the dynamic models of the robot are highly nonlinear or noises are nonGaussian. Then, the proposed SLAM framework is applied to high-frequency (HF) band radio-frequency identification (RFID) system. Since detection model of the HF band RFID system is highly nonGaussian, the traditional SLAM frameworks with the Gaussian assumption are not suitable and the proposed UPF-PS framework presents the superior performance. The proposed SLAM framework is evaluated through simulated experiments with an HF band RFID system and a range and bearing sensor. The comparisons between the proposed SLAM framework and four other SLAM frameworks, i.e., FastSLAM, UFastSLAM, and P-SLAM 1.0 and 2.0 are also given. Finally, the proposed SLAM framework is verified in a real experimental environment. The experimental results validate the effectiveness and superiority of the proposed SLAM framework.