Optimal Filtering Based Simultaneous Localization And Mapping

Positioning and navigation are two core research areas in autonomous mobile robot. There are a lot of mature methods to localize a robot in a known environment. However, the information of environment is usually unknown in many practical applications. So it is necessary to estimate the position of the robot only by using the relative observations and at the same time build the map of the environment. This process is called simultaneous localization and mapping (SLAM). Without global locator and human intervention, SLAM can enable the robot to navigate itself autonomously and map its environment simultaneously. Hence, SLAM is considered by many to be a key prerequisite of truly autonomous robots.This paper presents a brief study on optimal filtering based SLAM algorithms, and introduces some improved algorithms based on unscented Kalman filter (UKF) and H∞filter. The main contributions of this thesis can be summarized as follows:1. The kinematic model of robot, the model of environment, the model of noise, the model of sensor and the model of map are established in order to solve the SLAM problem. Moreover, the key technologies of SLAM are also analyzed.2. Concerning that it is necessarily to calculate sigma points by the square root of the state covariance matrix at each step in UKF based SLAM algorithm, this paper considers propagating the square root of the state covariance matrix directly in SLAM algorithm and proposes a square root unscented Kalman filter based SLAM algorithm.3. This paper proposes an extended H∞filter based SLAM algorithm. The proposed method requires no a priori knowledge of the noise statistics but assumes that the noises are bounded in certain energy level. Hence, this method is more robust than extended Kalman filter based SLAM algorithms.4. In order to extend H∞filter to nonlinear systems, this paper embeds unscented transform technique into the extended H∞filter structure and purposes an unscented H∞filter based SLAM algorithm.