Accurate localization for making Maps to mobile robots using Odometry and GPS without scan-matching

Masashi Yokozuka, Osamu Matsumoto

Research output: Contribution to journalArticlepeer-review

5 Citations (Scopus)


This paper studies an accurate localization method to make maps for mobile robots using odometry and a global positioning system (GPS) without scan matching. We investigate requirements for GPS accuracy in map-making. To generate accurate maps, SLAM techniques such as scan matching are used to obtain accurate positions. Scan matching is unstable, however, in complex environments and has a high computation cost. To avoid these problems, we studied accurate localization without scan matching. Loop closing is an important property in generating consistent maps. Inconsistencies in maps prevent correct routes to destinations from being generated. Basically, our method adds scan data to a map along a trajectory given by odometry. Odometry accumulates errors due, e.g., to wheel slippage or wheel diameter variations. To remove this accumulated error, we used bundle adjustment, introducing two types of processing. The first is a simple manual input moving a robot to a same position at start and end. This is equal that a robot returns to a start position at end. The second process uses a GPS device to improve map accuracy. Results of experiments showed that an accurate map is generated by using wheel-encoder odometry and a low-cost GPS device. Results were evaluated using a real-time kinematic (RTK) GPS device whose accuracy is within a few centimeters.

Original languageEnglish
Pages (from-to)410-418
Number of pages9
JournalJournal of Robotics and Mechatronics
Issue number4
Publication statusPublished - 2015
Externally publishedYes


  • Bundle adjustment
  • GPS
  • Localization
  • Mapping
  • Mobile robot

ASJC Scopus subject areas

  • Computer Science(all)
  • Electrical and Electronic Engineering


Dive into the research topics of 'Accurate localization for making Maps to mobile robots using Odometry and GPS without scan-matching'. Together they form a unique fingerprint.

Cite this