WARNING: HFO is not actively maintained! We accept pull requests and bug fixes, but please be prepared to do your own debugging in case of a problem. If you'd like to become a maintainer of HFO, get in touch.
Half Field Offense in RoboCup 2D Soccer is a subtask in RoboCup simulated soccer, modeling a situation in which the offense of one team has to get past the defense of the opposition in order to shoot goals. This repository offers the ability to quickly and easily interface your learning agent with the HFO domain. Interfaces are provided for C++ and Python.
- Boost-system, filesystem
- Qt4 [Required for SoccerWindow2 visualizer]: To not build the visualizer, add cmake flag
-DBUILD_SOCCERWINDOW=False
- Python 2.7 or 3.2+ (tested with 2.7 and 3.5)
- numpy
mkdir build && cd build
cmake -DCMAKE_BUILD_TYPE=RelwithDebInfo ..
make -j4
make install
Note: There is no need for 'sudo' with 'make install' - the files will be installed below the starting directory.
From the main HFO directory: pip install [--user] .
From the main HFO directory:
./example/passing_agents.sh
Start a 1v1 game played by Agent2D:
./bin/HFO --offense-npcs=1 --defense-npcs=1 --no-sync
Start an example agent on the empty goal task. This agent will move
forwards slowly. First start the server: ./bin/HFO --offense-agents=1 --no-sync &
and then connect the agent: ./example/hfo_example_agent
Or do both in a single command:
(./bin/HFO --offense-agents=1 --no-sync &) && ./example/hfo_example_agent
Example agents are provided in the example
directory. Below are two
minimal examples:
HFOEnvironment hfo;
hfo.connectToServer(...);
for (int episode=0; episode<5; episode++) {
status_t status = IN_GAME;
while (status == IN_GAME) {
const std::vector<float>& feature_vec = hfo.getState();
hfo.act(DASH, 20.0, 0.0);
status = hfo.step();
}
cout << "Episode " << episode << " ended";
}
hfo = hfo.HFOEnvironment()
hfo.connectServer(...)
for episode in range(5): # replace with xrange(5) for Python 2.X
status = IN_GAME
while status == IN_GAME:
features = hfo.getState()
hfo.act(DASH, 20.0, 0.0)
status = hfo.step()
print 'Episode', episode, 'ended'
The state and action spaces provided by the HFO domain are documented in the manual.