Hi everyone,
I'm looking for tips/advice on using accelerometer data to navigate a wheeled robot by dead-reckoning. I mounted an MPUninetwofifty on a small wheeled robot (similar to an RC car), and collected some data while the robot is moving. The graph below shows the accelerometer data (x-axis) from a run where I drove the robot forwards in a straight light, stopped, then drove backwards to the starting point. I can see the acceleration spikes (as expected) when the robot starts moving and when it stops moving, but there are also a lot of extra acceleration spikes during the entire time that the robot was moving. I guess these are probably picking up the vibrations from the motors and the wheels spinning?
Have any of you used the MPUninetwofifty for similar applications, and do you see similar results? Are there some good methods that work well to de-noise the accelerometer output to get more accurate acceleration data? I've heard about the Kalman filter, does it work well on data like this?
Thanks!
