240 likes | 373 Views
Autonomous Localization & Navigation using 2D Laser Scanners. Animesh Garg & Manohar Paluri. Outline. Problem Description Motivation Previous research Proposed approach Details of our approach Testing Results Conclusion. Why an autonomous painting system is required?. Introduction.
E N D
Autonomous Localization & Navigation using 2D Laser Scanners Animesh Garg & ManoharPaluri
Outline • Problem Description • Motivation • Previous research • Proposed approach • Details of our approach • Testing • Results • Conclusion
Introduction The Omnimove is a huge platform for moving very heavy weights around. Herein it would hold the robotic arm which will be used to carry out the painting job. During spray painting process, the environment has a very large concentration of paint particles decreasing visibility. And the paint settles on surfaces, it rules out markers based solution.
Potential Solutions • Cameras • GPS • INS • Sonar • Laser • Northstar • Vicon • And more…
Our solution • Laser Scanners • Cost effective, Reliable, Accurate, known solutions! • Paint Hangar constraints Sample Mount
Past Work • Fast RANSAC based registration algorithm for accurate navigation using only Lidar. RANSAC in combination with Huber's kernel to overcome the LIDAR input noise. • Hough transform for robot localization.The self localiza-tion technique in the paper is based on matching a geometric reference map with range information • RRT-Connect, bi-directional decision trees. • RRT* - Combines advantages of RRGs optimal solution with a tree structure.
Rigid Transformation • Scan1 Scan2 • Combined Scan
Line Extraction Techniques • Split-and-Merge • Line-Regression • Incremental • RANSAC • Hough-Transform • EM
Split & Merge • Initial: set s1 consists of N points. Put s1 in a list L 1 • Fit a line to the next set s in L 2 • Detect point P with maximum distance d to the line 3 • If d is less than a threshold, continue (go to 2) 4 • Otherwise, split s at P into s 1 and s 2, replace s in 5 • L by s 1 and s 2, continue (go to 2) • When all sets (segments) in L have been checked, 6 • merge collinear segments.
Hough Transform • Initial: A set of N points • Initialize the accumulator array (model space) • Construct values for the array • Choose the element with max. votes Vmax • If Vmax is less than a threshold, terminate • Otherwise, determine the inliers • Fit a line through the inliers and store the line • Remove the inliers from the set, goto 2
Find Maximas – 5 constraints • They appear in pairs: the first one is formed by peaks H1 and H2; the second one is formed by peaks H3 and H4. • Two peaks belonging to the same pair are symmetric with respect to the x-axis(angle). • The two pairs are separated by 90o • The heights of the two peaks within the same pair are exactly the same, and represent the length of the respective line segment. • The vertical distances between peaks within the pair are exactly the sides of the rectangle. In case of other obstacles in the scene, constraints 4 & 5 are not robust. So we only use 1, 2 & 3.