/swp-robot-runners

Primary LanguageC++MIT LicenseMIT

How to actually set up robot runners

Setup IDE

  1. Download Visual Studio Code or VSCodium
    • if VSCode runs inside of a Flatpak you need to add the following into settings.json under terminal.integrated.profiles.linux:
    "bash (outside flatpak)": {
    "path": "/usr/bin/flatpak-spawn",
    "icon": "terminal-bash",
    "args": [
        "--host",
        "--env=TERM=xterm-256color",
        "bash"
    ]
},
  2. Clone repository

Python setup

  1. create a venv with Python 3.10
  2. activate the venv
  3. python -m pip install -r requirements.txt
  4. pre-commit install to activate git hook that runs auto formatting and linting on commit

Install dependencies

Install dependencies on Ubuntu 22.04+ or Debian 11+:

sudo apt-get update
sudo apt-get install cmake build-essential libboost-all-dev python3-dev python3-pybind11

Install dependencies on Fedora 38+:

sudo dnf install cmake boost-devel python3.10-devel pybind11-devel

For all other Linux environments it's recommended to use Distrobox.

Compiling

  1. activate the python environment
  2. Using compile.sh:
./compile.sh

If compiling doesn't work try deleting the build directory.

Important commands

Run

in ./build:

./lifelong --inputFile ../example_problems/random.domain/random_20.json -o test.json

Run with benchmark tool

python3 benchmark.py --planner "astar" --iterations 200 --config '{"restarts": ["True", "False"]}'    

python3 benchmark.py --help
usage: benchmark.py [-h] [--rebuild] [--planner PLANNER] [--iterations [ITERATIONS]] [--config CONFIG]

optional arguments:
  -h, --help            show this help message and exit
  --rebuild             Use when you want to rebuild the program
  --planner PLANNER     name of the desired planner
  --iterations [ITERATIONS]
                        Specify the number of iterations(steps)
  --config CONFIG       Configurations to try out. Each possible combination of configs will be set as os environment variables and tested. - Use this to test out different parameters of your planner.

Benchmark tool will generate different plots in Output/ folder.

Visualize

  • clone PlanViz into the same directory as swp-robot-runners
  • visualize the run with PlanViz by providing the map and the generated "log" file test.json:
    • in ./build:
python3 ../../PlanViz/script/plan_viz.py --map ../example_problems/random.domain/maps/random-32-32-20.map --plan ./test.json --grid --aid --static --ca

Debug

It is not that easy to debug our Python MapfPlanner since it is called each second by the compiled c++ scirpt. Hence we need a remote-debugging server.

VS Code

import debugpy

Add the following lines:

# to wait for debugger connection:
debugpy.listen(('localhost', 5678))
print("Waiting for debugger attach")
debugpy.wait_for_client()
  1. call python3 benchmark.py or any other thing that calls our planner.
  2. In VsCode (while having pyMAPFPlanner.py open):
    1. Click on the Debug-Button
    2. Click "Run and Debug"
    3. Click "Remote Attach"
    4. Enter localhost as IP
    5. Enter 5678 as Port
  3. start debugging :)

IntelliJ

TLDR:

  • import pydevd_pycharm
  • Add line: pydevd_pycharm.settrace('localhost', port=12345, stdoutToServer=True, stderrToServer=True)

Original Readme:

Start-Kit

Join the competition

Log in to the competition website with a GitHub account, and we will automatically create a private GitHub submission repo for you. The repo will be the place where you submit codes. In the My Submission page, you can click "My Repo" to open your GitHub submission repo page.

Clone your submission repo

Clone your submission repo to your local machine. The repo contains starter codes to help you prepare your submission.

$ git clone git@github.com:your_submission_repo_address
$ cd your_submission_repo

Compile the start-kit

Dependencies

Install dependencies on Ubuntu or Debian Linux:

sudo apt-get update
sudo apt-get install build-essential libboost-all-dev python3-dev python3-pybind11

Homebrew is recomanded for installing dependencies on Mac OS.

Compiling

Using compile.sh:

./compile.sh

Using cmake:

mkdir build
cmake -B build ./ -DCMAKE_BUILD_TYPE=Release
make -C build -j

Run the start kit

Running the start-kit using commands:

./build/lifelong --inputFile the_input_file_name -o output_file_location

for example:

./build/lifelong --inputFile ./example_problems/random.domain/random_20.json -o test.json

more info on help:

./build/lifelong --help

Windows users

If you are a Windows user, the most straightforward method to utilize our start-kits is by employing the WSL (Windows Subsystem for Linux) subsystem. Follow these steps:

  1. Install WSL, please refer to https://learn.microsoft.com/en-us/windows/wsl/install
  2. Open a shell in WSL and execute the following commands to install the necessary tools (CMake, GCC, Boost, pip, Pybind11):
sudo apt-get update
sudo apt-get install cmake g++ libboost-all-dev python3-pip python3-pybind11
  1. Employ the commands provided above to compile the start-kit.

While it's technically possible to use our start-kit with Cygwin, Mingw, and MSVC, doing so would be more complex compared to using WSL. You would likely need to configure the environment yourself.

Upgrade Your Start-Kit

If your private start-kit copy repo was created before a start-kit upgrade, you could run the ./upgrade_start_kit.sh to upgrade your start-kit to the latest version.

You can check version.txt to know the current version of your start-kit.

The upgrade_start_kit.sh will check which file is marked as an upgrade needed and pull those files from the start-kit. It will pull and stage the files, but not commit them. This allows you to review the changes before committing them.

For files stated as unmodifiable in Parepare_Your_Planner.md, you always commit their changes.

The upgrade may overwrite some of your changes to CMakeLists.txt, compile.sh, and apt.txt, you could compare the difference using git diff and decide whether to revert some modifications or partially accept changes on these files.

The upgrade script will not touch any participants' created file, python/pyMAPFPlanner.py, inc/MAPFPlanner.h and src/MAPFPlanner.cpp. So that participants' implementations should not be influenced by the start-kit upgrade.

Input output description

Please refer to the Input_Output_Format.md.

Prepare Your Planner

Please refer to the Prepare_Your_Planner.md.

Debug and Visualise Your Planner

We provide a visualisation tool written in Python: https://github.com/MAPF-Competition/PlanViz. It is able to visualise the output of the start-kit program and help participants debug the implementations.

Please refer to the project website for more information. Also the document Debug_and_Visualise_Your_Planner which provides helpful hints for interpreting and diagnosing planner output.

Submission Instruction

Please refer to the Submission_Instruction.md.