By next competition, we hope to have a more accurate way of measuring the distance that Monty has traveled. This past competition, the only distance measurements we relied upon were those made by GPS; Although the calculations proved reliable to a certain extent, there is a lot of room for improvement. In order to increase both our accuracy and precision in determining where we are, we will be using a weighted average of GPS data and data from an encoder. Our weighted average algorithm is devilishly simple, but we needed to find the standard deviation of our GPS. On Sunday we sat Monty down and took 1,000 GPS points of the same spot, imported those points into Microsoft Excel, and took a look at the data.
Our findings were, to be honest, quite cool! First off, the standard deviation of both our latitude and longitude was .0579 meters, which means 95% of all of our GPS readings falls in a circle with a diameter of 11.5894964. That is insanely accurate! Of course, we had outliers on both ends of the data set and our range turned out to be 6.4 meters, but the incredible accuracy was pretty true to form for GPS.
Additionally, we took a look at the data in graphical form. The first graph we saw was a simple graph of Latitude vs Longitude. We predicted that this graph would have a bunch of data in roughly a circle in the center, with a few sporadic outliers dotted around the page. Much to our surprise, the bizarre output of Lat Vs Lon was very un-circle like, and had no mathematical shape that we know of (although it does produce a striking image of a roosted eagle).
The second two graphs we made helped us understand the output of the first graph a little better. We separately graphed Latitude and Longitude vs Time and once again our predictions were wrong. We expected another random dispersion of points along a best fit line, but instead we were greeted with two graphs that showed random downward and upward trends of points. We inferred that the rise and fall of the Latitude and Longitude were due to Satellite error or movement because the trends continue for a good time before switching direction. The line graph showed that the GPS doesn’t produce random points, but instead produces points that continue to increase (or decrease) in error IN THE SAME DIRECTION for a brief period of time.
RoboMagellan is scored by how long it takes a robot to get from the start to the final target cone, the faster the better. Times can be reduced by touching bonus cones, so the key decision is getting the best balance between the extra time it takes to get to bonus cones versus the reduction in time getting to those cones gives you.
