DEVELOPMENT AND IMPLEMENTATION OF OUTDOOR ENVIRONMENT MAPPING ALGORITHMS FOR MOBILE ROBOTS

Abstract

Range images are widely used in the fields of object modelling, medical imaging, mapping environments and mobile robotics. Most of these, require methods to combine two or more range images into a single model precisely. In particular, the domain of mobile robotics is revolutionised with the advent of range images. These act as primary input for obstacle detection, mapping, path planning and navigation algorithms. In this work, scan matching technique using Iterative Closest Point (I.C.P.) algorithm is used to model a real time 3D environment using multiple 2D laser scans. In this report, I.C.P. algorithm and its variants are explained. Instead of common three sigma filtering method (36) , two novel fast filters are proposed to refine and reduce data acquired. Data is reduced to 7.5% and more than 67% of singular points are removed. From results, it was proved that this little amount of data is enough for the mapping of the large environment. Different data structures and corresponding nearest neighbour search queries for the data structures are explained in detail. Two data structures, KD tree and Voronoi structure are implemented for the data storage and .a new hybrid structure is proposed for data handling. Different rigid body transformations are discussed in detail and Singular Value Decomposition (S.V.D) is used to derive orientation matrices. All the algorithms are. implemented in MATLAB 2009A environment. in addition to mapping, this algorithm can be applied directly to determine exact robot's position even in the situations (tunnels, underpass, and tall buildings) where Global Positioning System (G.P.S.) signal is