unhd (Roger Gonzalez ) (09/19/89)
My lab is involved in the design of experimental autonomous submersibles. In the past, we have used VMEbus based systems for high level control, and homebrew systems for the low levels. This has proven to be quite a bear development-wise, because we have to do all the multi-cpu work in software, in a system that was never designed for anything resembling our applications. I could bitch for hours and hours about it... I would like to see all the multi-cpu communication done in hardware, on a proven system on which we could use as much 'off the shelf' software as possible. (Right now, for example, we're stuck in the quagmire of getting a large unix lisp interpreter to run on a non-interpretative real-time OS that doesn't provide unix I/O.) Naturally enough, we started batting around the idea of using transputer boards. One of the employees here has a transputer board in his IBM clone, and he said that programming it was trivial and elegant compared to what we had been doing. He showed me a catalog of boards and software for it, and I was suitably impressed enough to start serious consideration of the idea. I need some good hard facts about whether or not this would work well. The scenario I envision looks like this: Inside the vehicle's computer tube, there would be an IBM PC bus, with PC motherboard, CGA card, and disk controller. The PC would be used for starting and stopping the vehicle, with the monitor and keyboard made easily removable. There would then be a 2 (?) transputer boards on the system, one of which would have serial I/O ports for interaction with sensors, propulsion, etc. Each CPU on the I/O board would have a low-level task program running, and each CPU on the second board would have high-level AI code running. Now.. please mail or post all comments, especially those pertaining to -Ease of development/programming tools -Real time uses of transputers -Existence of transputers with serial I/O ports -Prices of boards and software -Recommended products, dealers, and addresses. Thank you, Roger ------------------------- Roger Gonzalez Marine Systems Engineering Laboratory University of New Hampshire unhd!rg@cs.utexas.edu r_gonzalez@unhh.bitnet