Loading...

Combined Optical flow Algorithm and Simultaneous Localization and Mapping for Localization Robust to Surface Texture Variations

Takaloo, Saeed | 2017

746 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 49499 (08)
  4. University: Sharif University of Technology
  5. Department: Mechanical Engineering
  6. Advisor(s): Vosoughi, Gholamreza; Moradi, Hamed
  7. Abstract:
  8. The research field of localization with high precision (precision of 10 micrometers) has greatly attracted attentions in recent years. Optical flow sensor is one of the inexpensive solutions for the localization of robot, which have been used in computer mice. The most important application of this thesis is to localize walking micro robot. The main challenge of previous research is the fact that sensor's precision is highly dependent on the surface texture. The aim of this research is to implement an appropriate algorithm on optical flow sensor in order to increase the robustness of sensor's precision to surface texture. In this regard, firstly experimental hardware has been designed and assembled to give the image of the surface by optical flow sensor. Then, localization's setup has been utilized in order to move optical flow sensor on four surfaces (sheet of iron, granite stone, textile and typical paper) at four different angles (0°, 15°,30°,45°). Next, combined optical flow and SLAM algorithm have been implemented offline on the data obtained from experiment. In each surface, optimal values of the proposed algorithm's parameters have been calculated through Genetic Algorithm and the value of the parameters has been justified in accordance with surface textures. Zero degree and 30° paths have been chosen for optimization. Then, precision of displacement for the paths of 15° and 45° have been calculated. Results show that the proposed algorithm estimates the displacement on surface of iron with the precision of 443 µm without the need of calibrations. With calibration, localization's precision would be enhanced to 60 µm that is equal to sensor's resolution. Also, results show that utilization of proposed algorithm for the estimation of an angle has 5 degree errors on all surfaces. Finally we conclude that the combined template tracking and SLAM algorithm have a lower error in comparison with other optical flow algorithms, and the iron surface is the best choice for localization by the proposed algorithm
  9. Keywords:
  10. Localization ; Simultaneous Localization and Mapping (SLAM) ; Optical Flow ; Optical Flow Sensor ; Surface Texture

 Digital Object List

 Bookmark

No TOC