Effect of time-spatial size of motion image for localization by using the spotting method

Abstract

We have been developing a localization method for an autonomous robot without using any prior knowledge or assumptions about the environment. The method is general and robust and can be used in real world robot tasks. With a map description based on a reduced image sequence, we detect a frame sequence in the map which matches an input frame sequence, by means of reference interval-free continuous dynamic programming (RIFCDP). The method can cope with changes in the robot's speed because RIFCDP can spot arbitrary frame sequences by nonlinear matching. In this paper, we show a real-time localization system, and we illustrate the effectiveness of our method in the real world by some evaluation experiments.