I am working on a problem in robot tracking, where our main position sensor will be unavailable for up to a second due to its low frequency. That's why we would need to track its 2D-movements (x&y-axes) during these intervals using an accelerometer. I think the right term for this process is called dead reckoning. In case I lack additional sensors, I probably have to take the double integral of the acceleration...

What I am looking for, is a way to estimate the needed sensor-accuracy to achieve a defined tracking accuracy (say 10mm). Does maybe a closed form expression exist? Or is there any general work on this topic? (I couldn't find anything) The only method of doing this that I can think of would be simulating a test trajectory for the sensors and calculate the accuracy. But that would require to run a simulation of every possible sensor candidate...

  • $\begingroup$ Welcome to Robotics, Bobipuegi. This question looks like a duplicate of another relatively recent question here, please check out the answer there and edit this question to further clarify if/why that isn't sufficient. For your future reference, dead reckoning is using vehicle odometery to estimate location; i.e., using wheel encoders, not an accelerometer. I'm going to put the question on hold for now, but again if you still need help just edit this question and then flag it for moderator intervention and/or vote to reopen it. $\endgroup$
    – Chuck
    Commented Sep 24, 2018 at 17:56


Browse other questions tagged or ask your own question.