#TRIBBLE Accelerator Graph

Tribbles with Kirk from Star Trek

Today, I decided to tilt the accelerometer back and forth, like a rocking motion. Then I graphed the numbers to see if I could find a pattern. There were X, Y, and Z values. I only graphed the X and Y values, because the Z values seemed unpredictable, and with no visual pattern. Additionally, there were too many outliers to accurately graph all the Z values in a precise way. So, I graphed the X and Y values separately first, and then made a double line graph to see if there were any trends between the two values. I found that there was in fact a pattern for both the X and the Y values, but the two patterns did not match. With the exception of a few outliers, the X values went down about 2.5 IMUs and then went up about 2.5 IMUs and so on. The Y values were a little more scattered. They did indeed go up, down, and then up again, but the values were much closer together, and more spontaneous. First, they went down 0.5 IMUs, then they went up 0.35 IMUs. Next they went down 0.25 IMUs and then up 0.4 IMUs. Each repeated jump down following the previous values was about 0.15 more than the last, and the values got increasingly larger each time. Compared to the X values, they were a lot harder to predict. If I am going to try and make an alert of some kind happen when the tribble rocks back and forth, I will need to use elements of machine learning, and the final product may be unpredictable at best. I do not know for sure whether or not I will be willing to spend a lot of time on this one part. I do believe that if I can write code that will allow the tribble to accurately detect rocking 70% to 90% of the time, then it would be worth it. If not, then I will have to find another way to either detect the rocking, or I might not be able use the rocking motion as part of the final design. Next time, I will start to shape a polyester ball in order to make a shell that can separate the Raspberry Pi and battery, from the fur and possible external features.