Behaviour Tree

PyPi: https://pypi.org/project/py-bt/

Tree structure made up of Composite, Decorator and Leaf nodes.

Leaf nodes are where the behaviour happens, for example an "action" or "test".

Decorator nodes can only have one child and are used to wrap the result of this child node (e.g. "retry" or "not").

Composite nodes can have multiple children. Composite nodes can be of 2 types:

• Sequence
• Selector

Sequence nodes return the first failed child node. Similar to the ALL operator.

Selector nodes return the first successful child node. Similar to the OR operator.

Release History

• 1.2.1: Added tree model validation

• 1.2.0: Added the "retry" decorator node

• 1.1.0: Added the "not" decorator node (inverter)

Dependencies

Tested on:

• Python 3.8.1
• Python 3.7.5

Usage

Install:

pip install py-bt

Define a python module for you behaviour tasks (actions & tests). Tasks must return True or False depending on whether they succeed or fail. Tasks functions receive data and blackboard args. data is the input data to the tree execution and blackboard is a key value store where information can be shared between nodes. For example:

def choose_food(data, blackboard):
    if data["lunchbox"]:
        blackboard["choice"] = random.choice(data["lunchbox"])
        return True
    return False

def eat(data, blackboard):
    print(f"Eating {blackboard['choice']}")
    return True

Then, define your desired tree model in JSON or YAML format. For example:

{
  "tasks_path": "path.to.tasks.module",
  "tree": {
    "sequence": [
      {
        "task": "choose_food"
      },
      {
        "task": "eat"
      }
    ]
  }
}

Then initialise and execute a behaviour tree object with some input data:

from bt.behaviour_tree import BehaviourTree

data = {
    "lunchbox": ["apple", "orange", "pear"]
}

tree = BehaviourTree("/path/to/tree/model.json")
tree.load()

tree.execute(data)

Example Models

Some example models can be found under the /models directory.

For example /models/football/attacker.json contains a behaviour tree for how a attacking player in a football simulator might behave.

Tests

Run tests in local Python environment (use a virtualenv):

pip install -r requirements.txt -r requirements-test.txt
pytest
flake8

Run tests with tox:

pip install tox==3.14.5
tox

Upload to PyPi

rm -rf dist/
python3 setup.py sdist bdist_wheel
twine upload dist/*

References

• https://www.gamasutra.com/blogs/ChrisSimpson/20140717/221339/Behavior_trees_for_AI_How_they_work.php
• https://en.wikipedia.org/wiki/Behavior_tree_(artificial_intelligence,_robotics_and_control)
• https://py-trees.readthedocs.io/en/devel/