FRC Driver Station Logs in RapidMiner

Mike Young

Hello FIRST Participants Group - we are exploring the use of RapidMiner with Driver Station log files (.dslog). These log files are automatically generated and stored on the RoboRIO, so they are potentially something that every FRC team can easily access and use to get insights on their robot performance.

Background:

On-Robot Telemetry Recording Into Data Logs
https://docs.wpilib.org/en/stable/docs/software/telemetry/datalog.html

Good presentation on Data Visualization and Logging from FRC6328:
https://www.youtube.com/watch?v=fVLnJpItFZ8

Note that the .dslog files are binary, so you will need a utility to extract the data (.csv).  FRC6328 has developed AdvantageScope (Win/Mac/Linux):
https://github.com/Mechanical-Advantage/AdvantageScope/releases/tag/v2.2.2

There are other useful GUI clients and command line tools out there as well.

Attached are some test files in .dslog and .csv format
(.dslog source https://github.com/orangelight/DSLOG-Reader/tree/master/DSLOG-Reader%202/DSLOG-Reader-Tests/TestFiles )

The fields in the data are as follows:

• Timestamp
• /DSEvents
• /DSLog/BatteryVoltage
• /DSLog/CANUtilization
• /DSLog/PacketLoss
• /DSLog/PowerDistributionCurrents
• /DSLog/RioCPUUtilization
• /DSLog/Status/Brownout
• /DSLog/Status/DSDisabled
• /DSLog/Status/DSTeleop
• /DSLog/Status/RobotAuto
• /DSLog/Status/RobotDisabled
• /DSLog/Status/RobotTeleop
• /DSLog/Status/Watchdog
• /DSLog/TripTimeMS
• /DSLog/WifiDb
• /DSLog/WifiMb

We are just starting with RapidMiner, so any advice on how this data can be used and what insights can be found in the data would be very helpful!

Joseph Roy_20880

This exercise is definitely quite exciting. I have shared some slides with my observations especially related to outlier study with visualizations. Let me know, if this is interesting, so that we can focus on other attributes as well. Also, find attached the outlier data which was extracted using Automodel in RapidMiner. The data set under consideration is 2022_02_25 21_25_39 Fri-dslog.csv. 

Joseph Roy_20880

Please find attached the (.xml) process file of RapidMiner. Even though the process might seem bit complex, but it is quite simple. 

1. First step, would be to just connect Retrieve data to output.

2. Next step is cleaning data with turbo prep. Once, turbo prep is complete we need to just open the process in RapidMiner.

3. Eventually, after turbo prep the cleaned data is sent to Auto Model  for outlier detection. Below snapshot shows how you can directly go to Turbo Prep and Auto Model from Results window. Let me know if there is any issues.

 

Erik W. Larson

This is amazing work.