We present experimental results, obtained using a standard Polaroid sonar ranger and acoustic cats’ eyes, which show how substantial improvements in range and direction measurements are possible. The improvements in performance are dramatic and obvious.
A number of consequences follow from our approach. Firstly, in an unknown but designed environment, navigational data can be acquired in an
... [Show full abstract] initialisation phase that does not require the robot to move around. Secondly, once these beacon positions have been identified in the first scan, as the robot moves it need only direct its sensors to look in a small number of directions (equal in number to the number of geometric beacons being used to navigate) to correct errors that always arise when using odometry alone. Thirdly, mobile robot systems that need to move around to gather range data in order to disambiguate the spatially invariant reflections from the spatially transient specularities need highly accurate and precise odometry. Accurate precision odometry is needed because to correctly associate range measurements made to the same acoustical invariant, but from a different location, we need to know how the ranger has translated and rotated between these locations. Fourthly, odometry is no longer necessary, in principle, if a map is available. Fifthly, if a map and a position fixing system is available,1 and because we have high confidence in all our measurements, then when our measurements disagree with our expectations, we can be very confident that this is because there is an unexpected object in the robot’s task environment. This would be a useful competence for an industrial mobile robot.