Integrated PSO and line based representation approach for SLAM

This paper presents a novel method for integrating swarm intelligence and line-based representation of environment to solve the simultaneous localization and mapping (SLAM) problem of mobile robots. SLAM is a well-studied problem in mobile robotics. Because of stochastic nature of search strategy in swarm intelligence algorithms, they are very successful compared with other techniques in encountering SLAM problem. Line segment based representation of 2D maps is known to have advantages over raw point data or grid based representation gained from laser range scans. It contains higher geometric information that is closer to human insight and conceptual mapping, which is necessary for robust post processing. It also significantly reduces the memory and time complexity. Mobile robot reads raw laser sensor data in each step of its trajectory and converts it to a set of lines which is used to produce the last sensed map. At the next phase, the algorithm utilizes particle swarm optimization (PSO) and introduces a new evaluation function to find the actual state of the last sensed map inside a global map, which is merged into a global map by introducing a new merge method to reconstruct the global map. We use PSO's ability to run away from local extrema and converge towards an optimum point (i.e. best robot status in the map) by utilizing adaptive inertia weight strategy. We also introduce a new criterion to measure the similarity between the line pairs in the map. The experimental results on real datasets and virtual environments exhibit the algorithm's robustness, accuracy and superior performance on problems that are under consideration in SLAM such as loop closing, correspondence problem, curvature of the walls, and sensor uncertainty.