CPR ISS Division - Informatics

Informatics/data science is the field of study where data scientists draw useful conclusions from large and diverse data sets through exploration, prediction, and inferencing. While computer science mainly focuses on algorithms and programming, and statistics only studies mathematical principles and analysis, informatics has a broader focus extending to human behavior, big data analysis, cloud computing, web development, design and presentation, and more. Informatics is a perfect cross-section between technology and humanity.

At CPR, informatics helps us see hidden patterns and robot qualities that might not have been noticed.  We created a scantron-like “CPR Data Collection Interface” that enables us to utilize a fast, accurate, and tangible data input workflow.  To fully utilize the data we collect, we use Tableau Desktop® software to create elaborate visualizations.  Our Match Rubric and Match Strategy Sheet display an organized presentation of the upcoming match stats and enable us to better communicate with our robot operators and alliance members.  Our utilizations of informatics techniques give us a broader perspective and are imperative to our success at competitions.

Check out our free resources & tutorials at our Community Resources page

Quantitative Data Collection

CPR Data Collection Interface - Advanced Data Collection

Data is the essential foundation to any strategy and we had tackled both paper scouting and digital data collection via an app.  Both had their strengths—paper was tangible, and the app was automated and fast—but both had their downsides: paper needed to be manually inputed into excel, taking hours, and the app was prone to producing untraceable errors.  To achieve the best of both worlds, we created a scantron scouting system that integrated the best aspects of paper and digitalized scouting. If a scouter error was discovered, we could easily rectify the mistake by pulling out the physical scouting sheet for correction. Moreover, we would no longer need to painfully upload data inputs by hand.

So how does it work? Similarly to a standardizing test answer sheet, our scouters fill in answer bubbles on paper. These physical paper are quickly turned digital as we scan them with a student modified scanning app called ZipGrade. We find it amazing that, in an instant, random black dots from a “Scantron” scouting sheet can be automatically transformed into readable numbers and visualizations.

A Technical Explanation

Our “Data Collection System” is based on a highly modified ZipGrade vision target (ZipGrade 7913).  Initially, there were one hundred input groups (4 columns x 25 rows), each having 5 individual input bubbles.  This set up not only allows us to easily modify scouting sheet interface to suit different games, but also enables us to collect maximum data points (up to 500bit raw data) within a simple piece of 8.5x11 paper.  

Then, we place scouter interface layer right above ZipGrade vision target. During the design, being human-centric is our top priority. Instead of having scouters bubble in information such as match number, we automated a way to pre-fill this info via a printer. This improvement saved our scouting team a significant amount of time and prevented possible human errors.  We also created a Graphical User Interface to replace the old text based scouting sheet, this makes data recording much more intuitive and most scouters can easily use it without intensive training. Furthermore, encouraged by the FTC team 11120, we increased our font size by 36%.  This makes our “Scantron” scouting sheets much easier to read.

In addition to regular “Data Collection Interface” scouting sheets, we implemented “Data Verification” system at 2019 Houston Championship. This enables us to compare scouter data to official score on The Blue Alliance or FIRST Inspired websites. In the 2019 and 2020 season, we enlarged each input bubble by 31.6% and implemented an image threshold mechanism; this increased our scanning accuracy to 99.94%. What this means is that for every 1,600 bubbles or pieces of data that are scanned, there would only be one easy-to-fix error at most. We also replaced our old iPhone scanner with the document feeder scanner, this change enabled us easier and safer data access and shortened our data delay to only two matches.

“Mission Control” Data Management System - Instant Data Access

After “Scantron” scouting sheets are scanned, raw data is quickly imported to our "Mission Control” database.  By customized Excel functions such as “IF”, “VLOOKUP”, and “CHAR”, we are able to instantly decode (Base-6) data to conventional numerals for Tableau®, our data analysis software.  In addition to data export, we added even more features to assist our “Mission Control” manager.

Data Collection Error Detection

In addition to our old “sum based” (count bubbles) error detection calculations, we use logic analysis to eliminate impossible scouter feedback.  For an example, if a robot hangs during the endgame, then it should not be parking.  If a robot  does not hang, it should not be credited for parking… These feature helps “Mission Control” manager to find and eliminate unnecessary human errors. 

Data Verification System

CPR prides itself on collecting accurate data; therefore, we implemented “Data Verification” system.  This enables us to compare scouter data to official stats on The Blue Alliance or FIRST Inspired websites.  “Mission Control” managers can easily set up error tolerances or check data accuracy right at “Mission Control” dashboard. Data integrity is our top priority. Once an error is found, it is automatically filtered from Tableau until the error gets fixed or receives “ignore” approval. manager.

Feedback & Diagnosis Feature 

This Mission Control system enables our database manager to look up imported “Scantron” scouting sheets digitally without manually digging through giant folders to find hard copies.  This significantly shortened error-correction cycle time.

NAS Server and Real Time Data

By using a Network Attached Storage (NAS) server, several data analysts can simultaneously access our database. This way, everyone gets what they need in an instant.  NAS Server also enables everyone printer access without any complex setup.  Furthermore, this setup provides analysts and strategists wired internet access without having a wireless hotspot.                                                                                                                                                                                                                                                                       

Quantitative Data Analysis

Tableau® and Quantitative Data - Professional Data Analysis

To fully utilize the data we collect, we use Tableau Desktop® Software to organize it into elaborate visualizations. Tableau® helps our analysts see hidden patterns or robot qualities that might not have been noticed in the never-ending, black and white numbers of an excel sheet. The visuals we create in Tableau® are imperative to our success in matches, lobbying, and alliance selections. To accomplish all these tasks, we use stacked graphs, scatterplots, and match rubric dashboards.

Stacked Graphs

One visual we like to create in Tableau is the stacked bar graph. This graph allows us to see many variables of data on only one sheet, enabling us to see what a robot is capable of during a match. When making stacked bar graphs, we can also weight the variables, which highlights the qualities we value the most; this is similar to how a teacher may weight the test category at 50%, trumping the homework and classwork categories which could be set at 25% each.

Scatterplot Charts

Scatterplot Charts play a key role in preparing for alliance selections in the most logical manner possible. Scatterplots are one of the best ways to view many components of data on a single sheet. We use them to compare 2-4 robot qualities, helping us discover the most versatile robots in play. Last year, we valued robots who were strong in both the Scale and Opponent Switch; using scatterplots we could quickly discover which teams were excellent in both qualities.

Match Rubric Dashboards

Match rubrics are dashboards that we specially format to display nearly all the information that a match strategist needs to create smart strategies. Concisely displaying information on all three teams of both our own alliance and the opponent’s, a strategist can easily determine the most effective roles for our alliance partners by examining our team’s strengths and anticipating what our opponents’ aims are. Each match we play, we print out match rubrics customized to each new alliance that are given to our strategist and drive team. Having printed, trustworthy data on the spot helps the drive team build trust between the aligned teams as they prepare to work together.

Score predictions 

In our Match Rubrics, we created a new graph that analyzes the alliance teams’ past performances and then adds the three teams’ potential together to estimate the number of points they could earn in a particular match. Having this estimation set as a goal, our alliance can maximize their potential. We are excited to see how our strategist and drive team utilize this new tool.

Qualitative Data Analysis

The Second Lens to Analysis

Our team is committed to accurate and precise data; therefore, we use qualitative scouting to double check our data and observe robot qualities that aren’t caught by our quantitative scantron system. Qualitative data is essential to our team’s match strategies because it helps us better understand the strengths and weaknesses of alliance partners and opponents. For example, qualitative scouters can rate a robot’s aggressiveness to perhaps use them as a defensive bot, or if a robot constantly got stuck on the scale last year, we would consider placing them in a more fruitful position for points and not penalties.

Pit Scouting

Creating a Neighborly FIRST Community

A scouter’s job begins right when they enter the competition building. Pit scouting gives our team the opportunity to get to know our fellow teams and develop relationships. Simultaneously, pit scouting also allows us to gain information about other robots, including their drive train, speed, and overall capability. This task, typically assigned to new team members, introduces them to FIRST competition life and presents a valuable learning experience.