Learning Dynamic Cognitive Map with Autonomous Navigation

TL;DR

This paper presents a biologically inspired computational model for autonomous navigation that dynamically expands a cognitive map using active inference, enabling efficient exploration and rapid environmental structure learning without prior knowledge.

Contribution

Introduces a novel active inference-based model with a dynamically expanding cognitive map for autonomous navigation, inspired by animal strategies, improving adaptability and learning speed.

Findings

Model rapidly learns environmental structures within a single episode.

Achieves efficient exploration with minimal navigation overlap.

Operates without prior knowledge of environment dimensions.

Abstract

Inspired by animal navigation strategies, we introduce a novel computational model to navigate and map a space rooted in biologically inspired principles. Animals exhibit extraordinary navigation prowess, harnessing memory, imagination, and strategic decision-making to traverse complex and aliased environments adeptly. Our model aims to replicate these capabilities by incorporating a dynamically expanding cognitive map over predicted poses within an Active Inference framework, enhancing our agent's generative model plasticity to novelty and environmental changes. Through structure learning and active inference navigation, our model demonstrates efficient exploration and exploitation, dynamically expanding its model capacity in response to anticipated novel un-visited locations and updating the map given new evidence contradicting previous beliefs. Comparative analyses in mini-grid environments with the Clone-Structured Cognitive Graph model (CSCG), which shares similar objectives, highlight our model's ability to rapidly learn environmental structures within a single episode, with minimal navigation overlap. Our model achieves this without prior knowledge of observation and world dimensions, underscoring its robustness and efficacy in navigating intricate environments.