UPF를 기반의 다종 로봇 SLAM의 유효한 정보교환을 위한 연구

Abstract

Recently, research of autonomous mobile robots develops fast. The autonomous robots are adapted perform explore environment, search and rescue task. Truly autonomous mobile robots require the ability to map their environment using “Simultaneous Localization and Mapping” SLAM in order to explore it without getting lost. Multi-robot SLAM becomes necessary once an environment becomes too large. The thesis focuses on develop an algorithm to solve the multi robot SLAM problem. The Unscented Particle Filter will give us a better performance than Extended Kalman Filter. Thus we choose unscented particle filter to solve the problem. Thereafter we decompose the problem into two cases as the known initial poses case and unknown initial poses case to analysis. In the first case, we know the pose of each robot and in the second case, we assume the starting point is the origin of each robot frame, and one of the robots is the main robot. When the robots meet, they transform their information to the main robot’s frame to get a common map. Finally, we used two methods – EKF and UPF- to perform the multi-robot SLAM in simulation and compare the result of the simulation. The comparison can verify the method is efficienter than EKF method through simulation.