Autonomous Multi-Robot Mapping
Abstract
In this thesis a multi robot algorithm for Simultaneous Localization and Mapping (SLAM) using a Rao-Blackwellized particle filter was implemented. The particle filter was paired with a map-matching algorithm for calculating particle weights. The particle filter was then used to improve the map estimate of each robot exploring and mapping a labyrinth. The improved maps from each robot was then subjected to a map merging algorithm, which attempts to merge the partial maps from each robot together. The merged map will now be a better estimate of the real world, and remove the constraint that robots need prior knowledge of each others relative position.
The particle filter uses a motion model based on the robots odometry readings to estimate the current pose of the robot. Each particle will move according to the motion model, but will in addition incorporate some Gaussian noise. This will cause a distribution of belief states to appear, as each particle have a different trajectory over time and therefore produces their own version of the map. When a particle returns to a previously known position, a comparison between the current observations and the previous observation is performed using a map matching algorithm. The match between these map gives a score, also called weight, to the particle. The better match, the higher the score. Based on each particles score, it is chosen as a progeny, or discarded in a sequence called resampling. The particle with the highest score at the end of the exploration has a high likelihood of being a good representation of the robots true state.