JMaio/deliver.ai

Fine Tune the Accelerometer

Closed this issue · 2 comments

Look at how to detect what is abnormal

In the beginning, we were working with a simple thresholding technique.
Before this demo, we added a way of smoothing the input values by averaging over the last n values. The idea was to make the values a bit more stable and less susceptible to noise.
This averaged value is then thresholded in a similar way than before.
We still haven't achieved a satisfactory result.

Problem
the movement of the robot is quite jerky which causes the alarm to go off quite easily. We need to find a better way to distinguish the robot movement from manual movement.

New approach that still needs improvement can be found here -> 8f6f827