Mobile robot location identification system using reference interval-free continuous dynamic programming applied to time-varying images
Hiroshi Kojima, Yoshiaki Itoh, Ryuichi Oka · Systems and Computers in Japan · 1997
In this paper, we propose a method which enables an autonomous mobile robot to identify its location without using markers. In this method, time-varying images that the mobile robot acquires by walk-through are used as a reference pattern, and when the robot walks along a path that it has previously passed, the location can be identified. In order to realize this capability, reference interval-free continuous dynamic programming (RIFCDP) is applied. RIFCDP is an algorithm that detects an arbitrary section of the reference pattern which is similar to an arbitrary section of the input sequence, synchronized to each frame of the input sequence. Evaluation experiments have been performed in a real-world environment with the input sequences acquired by moving along the same course of the reference patterns, on a shifted course, and at different speeds. The results show that location identification is possible in all cases without mistakenly recognizing an unknown place as a known place. © 1998 Scripta Technica, Syst Comp Jpn, 28(11): 1–10, 1997