1. Objective
The objective of the robotics lab was to acquire an understanding of the accuracy and care it takes to gather legitimate data. We did this by calculating and analyzing the margin of error in a set of data.
2. Experiment Procedure
To perform our experiment, we needed to gather two sets of data to record the same information: one set recorded with natural human error, and a control set of data without any human error. For our sample data we recorded the distance traveled by a robotic car two different ways. The data with human error was gathered by means of our best judgement looking at the distance the car’s wheels had traveled along the side of a ruler. The control set with no human error was gathered using a computer program connected to the car. We used the formula Circumference=2(3.14)Radius to calculate and enter the distance one wheel rotation would make the car travel. The program used this information along with its sensory data of how many times the wheel turned during its trip to calculate for us the exact distance traveled by the car every time. We reset and recorded both numbers three times on the same setting of 100% power and 1 second of travel time. We then repeated the process with a different setting of 75% power and 1 second travel time. A third time, we recorded trips with a setting of 50% power and 2 seconds travel time. This gave us a total of 9 manually measured distance values that could be compared to 9 computer calculated distance values which are the following:
Measured: 37.5 36.9 35.9 24.5 25.4 26.0 32.6 32.3 34.0
Calculated: 39.5 39.5 37.6 26.1 26.1 26.1 32.4 33.2 33.5
We then took this data and compared it. We calculated the percent of error in each pair by using the formula provided in class. Our results are the following:
Percent error: 5.19% 6.8% 4.65% 6.8% 2.7% 0.38% 0.62% 2.83% 1.36%
3. Conclusions
Given the care we took to record the most accurate measurements that we could, we did not expect our percent error to be very large. At a maximum of 6.18%, I believe we succeeded in keeping our margin of error relatively small. What did strike me though, was the wide inconsistencies of error between trials. I thought our measurements had a consistent level of accuracy, but our error ranged from 0.38% to 6.8%. Although the scale of our measurements did keep our margin of error relatively small, the lack of consistency in error still showed me how much human error can warp data of a larger scale. This being known, machines are best to analyze large sets of data, and even so not all data can be trusted to be absolutely precise, because it is very hard to maintain both accuracy and consistency.