Practical Aspects of Total Least Squares Vectorization of Point Clouds in Mobile Robotics

Fast and reliable point cloud processing is a challenging task, especially when online running of the implementation on a mobile robot is required. This paper summarizes generally usable optimization techniques (hardware dependent implementation details are not covered) for vectorization of the point cloud using the least squares approach. Formulas for efficient implementation, methodology of tuning of the control variables, posprocessing for result reliability, as well as illustrative examples are all covered in the text. The discussed suggestions were experimentally proofed to give increased performance (in terms of speed and quality of approximation) with respect to basic implementation.