This is the companion repository to a paper presented at the International Workshop on Active Inference 2024 entitled:

"Planning to avoid ambiguous states through Gaussian approximations to nonlinear sensors in active inference agents."

The demonstrations folder contains notebooks with simple demonstrations of the model-predictive controller versus the free energy minimizing agents, as well as visualizations of the ambiguity landscape. The experiments folder details the Monte Carlo runs of the various free energy minimizing agents and the resulting visualizations.

Questions and comments can be addressed to the issues tracker.