Hippocampal formation-inspired probabilistic generative model.

Area of Science:

• Neuroscience and Artificial Intelligence (AI)

Background:

• Developing AI agents can be accelerated by referencing biological brain functions.
• Existing AI models often lack integration of neuroscientific principles.

Purpose of the Study:

• To propose a probabilistic generative model (PGM) for navigation in uncertain environments.
• To integrate neuroscientific knowledge of the hippocampal formation (HF) with robotics and AI engineering (SLAM).

Main Methods:

• Following the brain reference architecture (BRA) approach to compose the PGM.
• Surveying and discussing the relationship between HF findings and SLAM models.
• Designing the hippocampal formation-inspired probabilistic generative model (HF-PGM) for anatomical and functional consistency with the HF.

Main Results:

• The proposed HF-PGM integrates egocentric/allocentric information, drawing parallels with the entorhinal cortex to hippocampus pathway.
• The model utilizes discrete-event queues, consistent with neuroscientific observations.
• The study outlines methods for verifying the proposed model's efficacy.

Conclusions:

• Referencing brain function, specifically the HF, can significantly advance AI agent development.
• Integrating neuroscientific insights into AI models like SLAM offers a promising direction for creating more robust navigation systems.