On Vision-Based Orientation Method of a Robot Head in a Dark Cylindrical Pipe

This paper addresses the problem of navigation of an autonomous robot when it moves in a modern concrete sewer. The robot must keep its orientation within mostly cylindrical sewer pipes. This implies a geometrical constraint on the environment in which the robot operates. We present a hybrid vision system that consists of (a) an optical camera and (b) a laser crosshair projector generating a cross pattern in the camera field of view. The camera acquires the laser footprint projected on the pipe surface. The image of the footprint is the two intersecting curves. The shape of the curves depends on (up to the symmetry of the sewer pipe) a particular robot heading. We describe experiments conducted in the straight pipe segment of the dry sewer test-net. We present an algorithm that recovers instantaneous orientation of the laser projector relative to the pipe axis. We give a strategy for robot selforientation along the sewer pipe axis.