cphoenix@csli.Stanford.EDU (Chris Phoenix) (07/18/90)
I have a job in robotics research this summer, which covers everything from Action Networks to the hardware. My biggest headache at the moment is the sensors. We have a laser rangefinder and a ring of infra-red sensors, and all of them are quite noisy. Being a theoretical CS person, of course I don't have the knowledge to deal with imperfect hardware. So, any suggestions and/or references would be appreciated. The rangefinder projects a horizontal plane of laser light, and finds the range to the obstacle with a video camera mounted above the plane and looking down into it (the laser projects a line on an obstacle, and the farther up the picture the line is seen, the more distant the obstacle is). The IR sensors are just an LED and a photodiode. Both sensors are jittery, and the IR sensors also drift (as much as 30-50% in the raw data they return) over a period of minutes. The laser is too bright up close, and the line spreads out to cover quite a bit of the picture. So I can use suggestions or references on how to improve the hardware, or how to massage noisy data. Thanks! -- Chris Phoenix | "I've spent the last nine years structuring my cphoenix@csli.Stanford.EDU | life so that this couldn't happen." ...And I only kiss your shadow, I cannot see your hand, you're a stranger now unto me, lost in the dangling conversation, and the superficial sighs...
buck@drax.gsfc.nasa.gov (Loren (Buck) Buchanan) (07/19/90)
In article <14484@csli.Stanford.EDU> cphoenix@csli.stanford.edu (Chris Phoenix) writes: >So I can use suggestions or references on how to improve the hardware, or >how to massage noisy data. Thanks! Look at your basic signal processing texts as a starting point. (I'm also a CS type and have little relevant experience). I have a related question (which I am asking for others in my department), and that is on the subject of classifying the data returned from a sensor. We have a project to try and determine the cause of Gamma Ray Observatory going into "safe hold" mode. We get the recorded telemetry of some 200 sensors from the tape recorder dump and need to come up with classifications for the data to then feed into an expert system. Currently we are using a neural net to do this classification. My question is what should we consider to be noisy data? The other classifications we have are: rising, steady, falling, trough, and peak. What other classifications should we think about adding? Thanks & B Cing U Buck Loren Buchanan | buck@drax.gsfc.nasa.gov | #include <std_disclaimer.h> CSC, 1100 West St. | ...!ames!dftsrv!drax!buck | typedef int by Laurel, MD 20707 | (301) 497-2531 | void where_prohibited(by law){} CD International lists over 40,000 pop music CDs, collect the whole set.