MRSLAM - Multi-Robot Simultaneous Localization and Mapping

Download or read book MRSLAM - Multi-Robot Simultaneous Localization and Mapping written by João Alexandre Simões Martins and published by University of Coimbra. This book was released on 2013-09-01 with total page 67 pages. Available in PDF, EPUB and Kindle. Book excerpt: Nowadays, a collection of two or more autonomous mobile agents working together are denoted as teams or simply societies of mobile robots. In Multi-Robot Systems (MRS) robots are allowed to coordinate with each other in order to achieve a specic goal. In these systems, robots are far less capable as an entity, but the real power lies in the cooperation of the team. The simplicity of MRS has produced a wide set of applications such as in military tasks , searching for survivors in disaster hit areas, parallel and simultaneous transportations of vehicles and delivery of payloads. The success of single-robot Simultaneous Localization and Mapping (SLAM) in the past two decades has led to research on Multi-Robot Simultaneous Localization and Mapping (MRSLAM). A team of robots is able to map an unknown environment faster and more and reliably. However, MRSLAM raises several challenging problems, including map fusion, unknown robot poses and scalability issues. Rao-Blackwellized Particle Filters (RBPFs) have been demonstrated as an eective solution to the problem of single robot Simultaneous Localization and Mapping (SLAM), and a few extensions to teams of robots exist. However, these approaches are usually characterized by strict assumptions on both communication bandwidth and prior knowledge on relative poses between teammates. In this dissertation, we describe in detail a distributed MRSLAM approach using RBPF in the case of possibly constrained communication and unknown relative initial poses using Robot Operating System (ROS). We consider the environment as a two dimensional space with several obstacles, which are explored by a team of cooperative mobile robots, equipped with laser sensors. In order to eciently tackle the problem, the cooperation between agents and the memory space available for observations storage must be taken into account. Experimental results using a team of up to two robots in a large indoor area show the robustness and performance of the approach.