This repo contains discrete active inference for combining external goals and internal system requirements. The AiAgent class implements the methods for active inference while adaptive_action_selection implements the algorithm for conflicts resolution. The method parallel_action_selection instead, provides a list of possible alternative actions. Some lists contains multiple actions, which can then be executed in parallel. Every list in the list of possible actions is a different strategy.
The repo contains simple examples to use the class and the adaptive action selection algorithm.
If you find this work useful, please consider citing the connected paper:
- Pezzato, Corrado, Carlos Hernández Corbato, Stefan Bonhof, and Martijn Wisse. "Active inference and behavior trees for reactive action planning and execution in robotics." IEEE Transactions on Robotics 39, no. 2 (2023): 1050-1069
Install the package with:
pip3 install .or, for allowing local changes:
python3 -m pip install -e .Test an example script from the example folder:
python3 examples/example_parallel_act_sel.py