/rrc_phase2

Real Robot Challenge Phase 2 attempt - ThriftySnipe

Primary LanguagePythonBSD 3-Clause "New" or "Revised" LicenseBSD-3-Clause

Real Robot Challenge Phase 2 attempt - ThriftySnipe

Our report from phase 2 can be found here.

The code is built off the rrc_example_package provided by the challenge organisers. For more details on how to use this code with Singularity and ROS 2, see the relevant documentation.

Singularity Iimage

  1. Download our custom singularity image: user_image.sif. Otherwise, rebuild it yourself using 'user_image.def' and following these instructions.

  2. Name the image user_image.sif.

Train in Simulation

WARNING: Our method does not succesfully solve the Phase 2 rearrange dice task.

To train in simulation using domain randomization:

singularity run /path/to/user_image.sif mpirun -np 8 python3 train.py --exp-dir='dr_training' --n-epochs=200

This will start training with just 1 dice in the arena, and will add 4 dice every 50 epochs untill there are 9 dice in total. Details of all relevant arguments are found in rrc_example_package/her/arguments.py.

Real Robot

To deploy the trained model:

  1. Find the trained model at rrc_example_package/her/saved_models/dr_training/acmodel.pt.

  2. Upload the model to the robot cluster following these instructions. The path to the model on the cluster should thus be /userhome/acmodel.pt.

  3. Ensure your configuration file links to this repository.

  4. Login via ssh and call submit.

This should run the rrc_example_package/scripts/evaualte_stage2.py script on the real robot.



README from the original Example Package:

This is a basic example for a package that can be submitted to the robots of the Real Robot Challenge 2021.

It is a normal ROS2 Python package that can be build with colcon. However, there are a few special files in the root directory that are needed for running/evaluating your submissions. See the sections on the different challenge phases below for more on this.

This example uses purely Python, however, any package type that can be built by colcon is okay. So you can, for example, turn it into a CMake package if you want to build C++ code. For more information on this, see the ROS2 documentation.

Challenge Simulation Phase (Pre-Stage)

There are two example scripts using the simulation:

  • sim_move_up_and_down: Directly uses the TriFingerPlatform class to simply move the robot between two fixed positions. This is implemented in rrc_example_package/scripts/sim_move_up_and_down.py.

  • sim_trajectory_example_with_gym: Wraps the robot class in a Gym environment and uses that to run a dummy policy which simply points with one finger on the goal positions of the trajectory. This is implemented in rrc_example_package/scripts/sim_trajectory_example_with_gym.py.

To execute the examples, build the package and execute

ros2 run rrc_example_package <example_name>

For evaluation of the pre-stage of the challenge, the critical file is the evaluate_policy.py at the root directory of the package. This is what is going to be executed by rrc_evaluate_prestage.py (found in scripts/).

For more information, see the challenge documentation

evaluate_policy.py is only used for the simulation phase and not relevant anymore for the later phases that use the real robot.

Challenge Real Robot Phases (Stages 1 and 2)

For the challenge phases on the real robots, you need to provide the following files at the root directory of the package such that your jobs can executed on the robots:

  • run: Script that is executed when submitting the package to the robot. This can, for example, be a Python script or a symlink to a script somewhere else inside the repository. In the given example, it is a shell script running a Python script via ros2 run. This approach would also work for C++ executables. When executed, a JSON string encoding the goal is passed as argument (the exact structure of the goal depends on the current task).
  • goal.json: Optional. May contain a fixed goal (might be useful for testing/training). See the documentation of the challenge tasks for more details.

It is important that the run script is executable. For this, you need to do two things:

  1. Add a shebang line at the top of the file (e.g. #!/usr/bin/python3 when using Python or #!/bin/bash when using bash).
  2. Mark the file as executable (e.g. with chmod a+x run).

When inside of run you want to call another script using ros2 run (as it is done in this example), this other script needs to fulfil the same requirements.