Skip to content

Latest commit

 

History

History
95 lines (68 loc) · 4 KB

README.md

File metadata and controls

95 lines (68 loc) · 4 KB

Model Verifier

This folder contains the model_verifier.py controller which spawns automatically a robot and automatically searches for infringements to the rules or suspicious elements.

Keep in mind that this controller only checks part of the rules, therefore it's possible that a robot is invalid although it did not trigger errors nor warning during the analysis.

Dependencies

Additionally to Webots, you will need the python modules listed in requirements.txt.

pip3 install -r requirements.txt

Generating a pdf report requires the pandoc and texlive tools:

sudo apt install pandoc texlive

Robot inspection

To inspect a robot, you should make sure the following conditions are met:

  • The PROTO file of the robot is located in the robocup/protos folder or one of its subfolders
  • A file named postures.json which contains the upright and the extension poses is located in the same folder as the PROTO file

You can then run a model check using the following command:

./model_checker.sh <robot_name.proto>

This will produce the following result:

  1. Start a Webots simulation with an empty world
  2. Spawn the requested robot
  3. Open 2 windows allowing you to visualize the bounding objects, the posture of the robot and the solids annotations:
    • The first window will contain the upright posture, the second the extension posture
    • The following color code is used:
      • Brown: solids labeled as [arm]
      • Green: solids labeled as [hand]
      • Black: solids labeled as [foot]
      • Grey: unlabelled solids
    • Once you close a window, the verification process will continue
    • It is possible to skip this step using the option --no-display
  4. The simulation will close automatically and the results will be available in the folder results/robot_name:
    • model_verifier.log contains the messages from the model_verifier controller
    • simulator.log contains the standard output of the siimulator
    • simulator.err contains simulator errors and warning raised when loading the model
    • robot_properties.json A file that can be used to import robot dimensions on the submission website
    • report.pdf contains an overview of all the results including:
      • Warnings and Errors during model_verification.
      • Table with all the joints and their values
      • Table with the main properties of the robot
      • Logs of the model verification
    • report.md the source file for report.pdf

Archive inspection

To inspect the validity of an archive before uploading it on https://submission.robocuphumanoid.com you can run the following command:

./archive_checker.sh <archive.zip>

It will check some properties of the archive, extract its content in the robocup/protos/tmp folder and then perform an automated inspection with the flag --no-display.

Run mock game

This script is used to detect low performances for the robot and currently requires to perform some manual step.

  • The game file game.json in the referee controller has to be configured as follows:
    • "minimum_real_time_factor" must be set to 1
    • "max_duration" must be set to 120

This script takes as argument the path of a folder team_name which has to be organized in the following way:

team_name
|- team.json
|- robot_name
|  |- posture.json
|  |- robot_name.proto
|  |- ...

Game can then be launched with ,/run_mock_game.sh team_name or ./run_mock_game.sh --no-rendering team_name:

This will start a game with a full team red and a full team blue, both of them will have clients connecting after a little while in order to ensure robots are not in asleep mode.

The output of the clients will be stored in different files in folder team_name (e.g. RED_1_with_rendering.txt). At the end of the test, robot is closed and the last output from the clients are shown on screen.

The value obtained can be used to estimate the speed factor associated to the team.