Game AI algorithms in pure Python
Project description
pychology
Simple Python implementations of popular game AI techniques.
Current State: Very alpha. Only behavior trees are implemented.
Behavior Trees
Behavior trees are tree structures that represent the logic of choosing and switching the current task. Each node in the tree can be considered as a function that is called, and returns an indication of the current task's state; Whether it is active, done, or has failed.
The tasks themselves are the leafs of the tree. pychology
expects you
to provide functions to call when a task is started or continued, and
that it returns one of the above mentioned node states, namely
NodeState.ACTIVE
, NodeState.DONE
, or NodeState.FAILED
. These
functions then are plugged into pychology.behavior_trees.Action
.
Inner nodes either have a single child and do some logic; These are called decorators, and do a lot of heavy lifting. They can change the returning node state, prevent sub-trees from being executed, check for conditions, count how often or for how long a sub-tree has been active, provide debug information, and much more.
There's also nodes that have multiple children. The two most popular are the Chain (usually called Sequence), which executes a series of tasks (or rather sub-trees) one after the other, and Priorities (also called Selector or Fallback) which tries running one sub-tree after the other until one reports something other than failure.
The class pychology.behavior_trees.BehaviorTree
can act as a root node
that automatically resets the tree's state after it has finished or
failed, and can also run arbitrary code to integrate with the wider
environment that the tree is embedded in.
An example on how to go through a door:
BehaviorTree(
Chain(
DoneOnPrecondition( # The first task is done
is_door_open, # if the door is open,
Priorities( # otherwise try to
Action(open_door), # open it
ReturnActiveOnDone( # or try again after
Action(unlock_door), # unlocking it,
),
ReturnActiveOnDone( # or try again after
Action(pick_door_lock), # picking the lock,
),
FailOnPrecounter( # or try
4, # three times to
Action(kick_in_door), # kick in the door.
),
),
),
Action(walk_through_door), # Afterwards, go through.
),
)
TODO
- Project
- Packaging
- Behavior Trees
- Multinodes: Parallel, weighted random choice
- Decorators: More return value logic
- Blackboard support
- Debug tools
- Visualization
- De-/Serialization
- Tests
- Hierarchical Finite State Machines (HFSM): Everything
- Goal-Oriented Action Planning (GOAP): Everything
- Hierarchical Task Planning (HTN): Everything
- Whatever this is
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distributions
Built Distribution
Hashes for pychology-0.1a0-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 7b70fd245101a91ff78c41ca270ce0ccbeeacf66e1c6c947a2cd7adb00ac3d64 |
|
MD5 | d6b72fd30d922181aab088b46679efef |
|
BLAKE2b-256 | c60d0d8cdd9da046419e5b02ef6325b5c323dc4381cc87dec3dffa597b502ec3 |