With that added functionality, it looks like I'll have to buy a second robot, cause that would pretty much fill all my requirements, but I don't want to abandon my current project (which is alreayd based on the robotix base).
Now to digress: One of the things I have realized from tinkering with this is that true stereo vision is not possible with the SRF sensors currently produced. If a ping could be sent out, and a pair of sensors detect the resulting echos, a comparison of the timing of the echos could locate the object relative to the sensors. In my case, I have a pair of SRF08 sensors, and compare the echos from the pair to locate an object, but they are detecting differnet pings, rather than the same ping. This introduces more error than is strictly necessary.
If there is a sensor out there with one emitter, and two receivers with sufficient separation between them that a direction can be identified, I have not seen it. Technically, it wouldn't be all that much different from the SRF08 sensor. If the return from the sensor was a pair of time difference from the ping to detection for the two sensors, the math would be straightforward, and could be done off the chip.
This is an interesting idea (regarding your dual receiver idea). I would expect, however, that you would need a wide beam pattern for this to work, and it seems that the general trend in ultrasonic sensor design is to make the beam as narrow as possible to increase the resolution of the detection grid. This is great if you have multiple sensors mounted at different angles which are activated in sequence, or if you mount the sensor to a panning head. I would guess that this is precisely what you are trying to avoid.
I don't have any ideas, except to recommend that you study the Maxbotix product line. Their sensors provide more options than the Devantech sensors, and they might have something with the necessary configurability to achieve the beam width you need.
James Y. Wilson http://www.learningce.com Fortudine Vincimus!