Error Eliminating Rapid Ultrasonic Firing (EERUF -- pronounced "e-rough") aims at reducing range-errors due to ultrasonic noise and to eliminate crosstalk. By doing so, EERUF allows fast, asynchronous sampling of sensors, typically on the order of 60-100 ms for 8 - 12 sonars .
Ultrasonic sensors are subject to noise and sporadic false readings, just like any other sensor system. In most operating environments (except for some shop floors), environmental ultrasonic noise is fairly rare. However, robots with multiple ultrasonic sensors may introduce their own noise, a phenomena known as crosstalk. Crosstalk differs from other environmental noise because it introduces a systematic error that will repeatedly cause similar erroneous reading (see Fig. 1). A related kind of "semi-systematic" error is crosstalk that may occur when multiple mobile robots with multiple ultrasonic sensors operate in the same environment. In most indoor applications, crosstalk is much more likely to occur than environmental ultrasonic noise.
Fig. 1: How crosstalk is generated
At an instance t_0, sensor x is fired and its soundwaves are reflected off three walls. Assuming the walls are fairly smooth (e.g., drywall), the reflected wavefront will reach sensor y after traveling through the distance L=l_1+l_2+l_3+l_4. If, at this time, sensor y is awaiting an echo of its own, then it will receive the signal from sensor x and interpret it as its own echo.
When operating a mobile robot under such real-world conditions, it is unpractical to base the decision for an obstacle maneuver on a single (possibly erroneous) sensor reading that seems to indicate the presence of an object in front of the robot. A more practical approach requires multiple range samples to be combined in order to increase the "confidence" of the algorithm in the existence of an obstacle, i.e., the signal-to-noise ratio. When traveling at higher speed (e.g., V>0.3m/sec), it is crucial to quickly and repeatedly sample each sensor, to gather multiple samples in time to avoid a collision. Fast sampling of multiple sensors, however, introduces even more ultrasonic noise and increases the occurrence rate of crosstalk.
To overcome these conflicting conditions, we have developed a method that allows rapid sampling of all sensors while reducing erroneous readings due to environmental noise by a factor of 10-30 (depending on the application). Moreover, our method completely eliminates crosstalk produced by either on-board sensors or sensors from other robots operating in the same environment. EERUF can be implemented in software and works with the off-the-shelf ultrasonic sensor interface board and sensors from POLAROID.
Fig. 2: Using EERUF with 8 ultrasonic sensors, a TRC LabMate zaps through an obstacle course of densely spaced, 8 mm thin wooden sticks. The maximum speed was 1 m/sec in this run.
For more details on EERUF see papers 32 and 33
This file last updated on 7/4/96 by Johann Borenstein email@example.com