Shigeru Bando

Use of the parallel and perpendicular characteristics of building shape for indoor map making and positioning

We are trying to develop an easy autonomous map making method for indoor mobile robots, using the property of the usual buildings, which we can assume that the major walls are arranged parallel or perpendicular. Based on these properties, the robot can find its own direction easily, with its surroundings range data, so that the accumulated error in the calculated direction can be avoided. In this paper, we propose a method to know the robot direction using scanned range data of the environment walls, and the method to build the map by connecting the scanned range data using this property. We also present an example of experimental results of indoor map making by a mobile robot without using the odometry, which shows the effectiveness of the proposed method.

Scan matching method using projection in dominant direction of indoor environment

In this paper, we propose a new scan matching method, which uses the feature of structure in an artificial environment where walls are located parallel or perpendicular to one another. This method can precisely obtain the relative pose between two scan data using the dominant direction with a small calculation cost. Efficacy of this method is proved by an experiment conducted in an indoor environment. Graphical Abstract

Global localization of a mobile robot in indoor environment using special frequency analysis of 2D range data

We propose a new approach for finding a location when a mobile robot stays in a indoor environment. In this approach, a spatial frequency analysis is applied to 2D range data obtained by a laser scanner, which is conducted as follows: first, a mobile robot surveys the whole subject area, collecting 2D range data consecutively at short intervals. The collected 2D range data are used as a reference. Second, the robot is placed somewhere in the environment, then it takes current 2D range data at the place, and finds its location out in the reference data. All the 2D range data are treated as binary images and are transformed into the spatial frequency domain. Matching between the reference and current 2D range data is conducted on the spatial frequency domain. This method has an advantage of exploiting the cross-correlation index for similarity evaluation of the matching.

Road following based navigation in park and pedestrian street by detecting orientation and finding intersection

Technologies of autonomous outdoor navigation are required for automatic porterage, transportation, and service robots systems. People can go through the town to get to desired building without using global positioning system. He walks by detecting his orientation and knows which street is on. And he finds the intersection and turns. In this paper, we propose the outdoor navigation method based on the road following with detecting orientation and finding intersection, which can work on the real world. We use two scanning laser range sensors, for obstacle avoidance and finding the edge of the road, and a magnetic field sensor. We describe the implementation of the system. And we show the experimental results of the system in park and pedestrian street of Tsukuba city.