ICE matching, robust and fast feature-based scan matching for an online operation
There are three challenges in finding robot location and environment map which are accuracy, robustness and computational cost. Some simultaneous localisation and mapping (SLAM) methods work perfectly with ideal environment without uncertainties, noise and disturbances; however, in the real world, the performance considerably decreases. Designing a practical method of finding the robot position during motion, particularly when it is fully autonomous, should not be time consuming. This point has been forgotten in many SLAM approaches. Finding a comprehensive algorithm for solving all the problems is very complex. In this paper, with a new geometrical viewpoint, the objective is to solve them simultaneously as much as possible. However, with the proposed technique, sub-optimal solutions of different targets will be found. This method, named ICE matching (Intersection, Corner and End of Wall features), is capable of being combined with different mapping algorithms to solve the SLAM problem. Defining new informative features and novel matching and hierarchical optimisation mechanisms, congregated in this method, create a robust practical technique in terms of accuracy and convergence rate. To evaluate the accuracy, qualities of different popular SLAM algorithms are compared with the present approach in some similar data sets. In addition, evaluating ICE matching on a real robot with manual and autonomous driving shows its considerably high performance and robustness against noises and shortage of data in online applications, in comparison with other approaches.
No Reference information available - sign in for access.
No Citation information available - sign in for access.
No Supplementary Data.
No Article Media
Document Type: Research Article
Affiliations: Faculty of Electrical, Computer and IT Engineering, Qazvin Branch, Islamic Azad University, Qazvin, Iran
Publication date: March 4, 2015