GOAPLite

by Matt Young, 2020.

This is my implementation of Goal Oriented Action Planning (GOAP), a simplified STRIPS-like planning algorithm commonly use in games. GOAP allows agents to intelligently make long-term decisions based on a description of the actions they can perform, and a set of boolean variables describing their environment.

GOAPLite has been specifically designed for use in embedded systems, specifically our RoboCup Jr Open Soccer robotics team. With that in mind, it's written in pure C11 and designed with minimal overhead, simplicity and future-proofing in mind.

Currently, the planner uses a simple and inefficient depth first search algorithm. In future it could be rewritten to use A* or one of its variants.

Actions are currently loaded via a JSON file for ease of debugging, however, any other format such as Protocol Buffers or a custom format could easily be added.

Partially inspired by this library: https://github.com/cpowell/cppGOAP

Dependencies

All dependencies are bundled with the repository in the lib folder, unfortunately there are some. However, they are all MIT licensed (cJSON, map) or public domain (DG_dynarr) so it shouldn't cause any licensing dillemas.

• CJSON to parse action description JSON documents. Can be removed if you have another way to declare these.
• rxi's map to store world state. Essential.
• DG_dynarr for linked list implementation. Essential.

In future, I'm going to aim to implement these dependencies in GOAP code itself so you don't have to worry about dependencies.

Compiling and running

This is a CMake project, so it should just be a matter of cloning the repo and doing the usual:

mkdir build
cd build
cmake-gui ..
make -j 4
./main

(or something along those lines).

The project is developed using CLion, so you can also just import it as a CLion project and run main.c

Integrating in your own project

The way I would recommend integrating this into your own projects is simply copying goap.c, goap.h and the dependencies in the lib folder somewhere into your source tree. It may be possible to build this as a shared or static library, but it's untested.

Once that's done, you should read main.c for a usage example on how to load an action list, set up a world state and solve the problem.

GOAP resources

• https://gamedevelopment.tutsplus.com/tutorials/goal-oriented-action-planning-for-a-smarter-ai--cms-20793
• http://alumni.media.mit.edu/~jorkin/goap.html
• https://en.wikipedia.org/wiki/Stanford_Research_Institute_Problem_Solver
• The PDFs in the docs folder in this repository

Final notes

More documentation would be ideal, I agree. Especially Doxygen stuff, I'm going to get onto that one day.

Unfortunately, this project has come to somewhat of a standstill due to the fact that it's not really being used for anything (putting it on our RoboCup Jr robot has sort of been stalled). If anyone has a use case for it, please let me know with a GitHub issue or email and I'd be more than happy to open it again.

Good luck!