抄録
The Mechanical Engineering Laboratory started a six-year project of Guide Dog Robot, which we call MELDOG after our laboratory, in the fiscal year of 1977. This project researches fundamental control and communication problems of a man-machine system, i.e. (1) how the robot guides itself by using landmarks set in the town and an organized map of the landmarks and how a blind traveller follows the robot (control of a man-machine system) and (2) how the blind person can find obstacles making use of the visual and ultrasonic sensors on board the robot (obstacle detection by man-machine cooperation). In previous reports a method for guiding a robot vehicle using discrete landmarks (white painted lines on the streets with a length of 2m and a width of 0.15m) and an organized map on the connection of landmarks stored in the memory of the robot was proposed, and its feasibility was demonstrated by both computer simulation and an outdoor experiment with the test hardware MELDOG MARK II. The robot traveled from one landmark to another, setting its steering angle at the value previously stored in the memory of the robot. When it reached a landmark, it adjusted its orientation and position by following the mark using the landmark sensors. Because of this adjustment of position and orientation at each landmark, course error was cancelled at each landmark. It was necessary, however, to set landmarks every ten meters to ensure the guidance of the robot. In order to make the robot travel longer without using the landmarks set on the road, it is necessary for the robot to use natural scenery as an auxiliary landmark. In this report, a method of course control of a travel robot using a direction-controlled visual sensor is proposed, and its feasibility is demonstrated by both simulation study and preliminary experiments with the test hardware. The direction-controlled visual sensor is a micro-computer-controlled CCD camera with a posture-controlled mirror attached, all mounted on a turntable. The mirror is controlled so that the robot can see far and near, and the turntable is controlled so that it can see forward, to the left and right side. The robot detects the edge of the street using the direction-controlled visual sensor, and the course of the travel robot is controlled by the algorithm proposed by using the information about the street.
