This repository is the implementation of the paper "MGCBS: An Optimal and Efficient Algorithm for Solving Multi-Goal Multi-Agent Path Finding Problem"

This repository is based on code from libMultiRobotPlanning.

The full dataset can be downloaded here.

The dependency includes:

• yaml-cpp
• Boost
• PkgConfig

A small dataset is given in ./dataset

To build the project, using the following instructions

mkdir build
cd build
cmake ..
make

The executable is generated in ./build/main

To run the executable, the input file and the output file should be given.

Go to the root directory of the repository.

cd ..

Use the following instruction to run the executable.

./build/main -a <algorithm> -i <input_file> -o <output_file>

The <algorithm> can be:

• MGCBS (Corresponding th 'MGCBS (A3)' in the paper)
• MGCBS_w1 (Corresponding th 'MGCBS without TIS Forest (A2)' in the paper)
• MGCBS_w2 (Corresponding th 'HCBS (A1)' in the paper)
• OptimalCBS (Corresponding th 'CBS + A* (A4)' in the paper)

For example

./build/main -a MGCBS -i dataset_example/lak303d/2_4_0.yaml -o output.yaml

In the name of input data file "x_y_z.yaml", 'x' is the agent number, 'y' is the goal number and 'z' is the instance id.