Autonomous Mobile Robot Map Building, Exploration And Localization

This paper researches on Simultaneous Localization and Map Building (SLAM) and Map Exploration (ME) of Autonomous Mobile Robots(AMR) based on uncertainty environmental information, in order to build a complete system that makes robots have the ability of autonomous map development in complete unknown structural environment.First, the paper compares different range sensors, analyses infection on laser range finder of various environmental factors and shows the advantages of laser range finder in mobile robot research. The paper concepts the uncertainty of distance and angles of laser range finder, builds error model of laser range finder and odometer, and designs architecture of SLAM and ME using layer structure from up to down.Then, after backward glance and summary of methods on mobile robot SLAM, this paper proposes a new there-step geometry map building approach based on probability model: improved angle histogram, weighted least square fitting and dynamic feature updating. Calculation of line feature's covariance matrix realizes discretization, so decreases the memory and compute burden. The paper also realizes precise localization using Extended Kalman Filter (EKF), and applies localization results to map building to form complete SLAM system. The experimental results demonstrate the accuracy, real-time, and robustness of SLAM system, and fulfills the design requires completely.In ME research, the paper expounds the difference between mobile robot map exploration based on geometry map and other navigation methods firstly, then designs the three layer architecture of ME separately: global goal planning, local path planning and real-time obstacle avoidance. In global goal planning, proposes the unknown field concept to search goal planning on high layer and uses decision tree data structure; in local path planning, plans the path composed with lines and arcs using sub-goal and arc track method, and validates path safety using rectangle safe field; In real-time obstacle avoidance, conforms virtual avoid points with raw sensor data, and plans the path with arc tracks in order to assure the safe of robot in dynamic environment.