Simultaneous localization and mapping algorithms with environmental-structure prediction for single- and multi-robot systems

Her-Jye Chang, Purdue University

Abstract

A major goal in mobile robotics is to develop autonomous robots that can perform various tasks in unknown environments without human guidance. To achieve this goal, a mobile robot must have the capability to sense and interpret its surroundings, and localize itself in the environment. This capability is usually called Simultaneous-Localization And Mapping (SLAM). The SLAM problem is difficult because the mapping and localization processes are intertwined, making it a complex search problem in a high-dimensional space. This thesis focuses on fundamental research that explores and investigates efficient single- and multi-robot SLAM algorithms. For the single-robot SLAM problem, we propose a novel prediction-based SLAM algorithm, called P-SLAM. Unlike traditional SLAM algorithms, the proposed P-SLAM predicts and utilizes repeated structures/features in an environment to facilitate the search process. The prediction is based on the observation of the surroundings of an unexplored region and comparing it with the built map of explored regions. If a match is found, a hypothesis is generated to indicate that a similar structure has been explored before. The mobile robot can utilize the predicted structure as a virtual mapping, and decide whether or not to explore the unexplored region to save the exploration time. If the mobile robot decides to explore the unexplored region, a correct prediction can be utilized to reduce the high-dimensional search space in the SLAM problem and speed up the SLAM process. Furthermore, we derive the Bayesian formulation of the proposed P-SLAM, which integrates the predictions into a probabilistic-SLAM framework. This allows us to implement the proposed P-SLAM with an existing SLAM algorithm. A multi-robot SLAM algorithm (MR-SLAM) is expected to provide better efficiency, accuracy and reliability than a single-robot SLAM algorithm. However, most of existing MR-SLAM algorithms focus on map fusion and discard the scalability issue of environmental size and the number of robots. We propose and develop a scalable multi-robot SLAM algorithm with topological and metric maps, called MRTM-SLAM. This algorithm builds a graph-like topological map with vertices representing local metric maps and edges describing relative poses of adjacent local maps. By dividing a SLAM problem into several smaller SLAM problems, the proposed MRTM-SLAM algorithm greatly reduces computational requirements and provides the scalability in the size of an environment. In addition, inserting an edge between their topological maps can naturally do the map fusion between two robots. Thus, the proposed MRTM-SLAM algorithm is also scalable in the number of robots. Extensive computer simulations and experiments have been conducted on Pioneer-3 DX mobile robots to validate the performance of the proposed P-SLAM and MRTM-SLAM algorithms.

Degree

Ph.D.

Advisors

Lee, Purdue University.

Subject Area

Electrical engineering|Robotics|Computer science

COinS