Free-Energy-Lutar Active Inference with Hierarchical Predictive Coding and Cognitive Map Navigation

Key Points

• The aim is to present a cognitive architecture integrating active inference and hierarchical predictive coding for better navigation and decision-making.
• Introduces Free-Energy-Lutar as a variational free energy minimization framework.
• Describes Predictive Coding Error Minimizer incorporating Rao-Ballard hierarchical predictions.
• Presents Cognitive Map Navigator modeled on grid-cell navigation.
• Completes the cognitive architecture covering aspects like trust, quality, retrieval, safety, and memory.
• Validates the framework through a TypeScript implementation.