Zero-shot Structure Learning and Planning for Autonomous Robot Navigation using Active Inference

2025-10-10

AI Summary

To address autonomous navigation in unknown environments without prior maps or large-scale labeled data, this paper proposes AIMAPP—a unified framework grounded in active inference. AIMAPP integrates topological mapping, biologically inspired place-cell encoding, and episodic memory to construct sparse topological graphs online and enable dynamic path planning. Its generative model, driven by expected free energy minimization, supports zero-shot transfer, fully self-supervised learning, and drift-resilient exploration. Crucially, AIMAPP embeds neurobiologically plausible mechanisms—such as hippocampal-like spatial coding—within the active inference paradigm, thereby enhancing both localization robustness and decision interpretability. The system is ROS-compatible and hardware-agnostic. Extensive evaluation across large-scale real-world and simulated environments demonstrates that AIMAPP significantly outperforms state-of-the-art methods under sensor noise, environmental dynamics, and perceptual ambiguity.

Technology Category

• Intelligent Robots: Localization, Mapping, and Navigation
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Planning, Routing, and Scheduling: Learning for Planning and Scheduling

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Autonomous navigation in unfamiliar environments requires robots to simultaneously explore, localise, and plan under uncertainty, without relying on predefined maps or extensive training. We present a biologically inspired, Active Inference-based framework, Active Inference MAPping and Planning (AIMAPP). This model unifies mapping, localisation, and decision-making within a single generative model. Inspired by hippocampal navigation, it uses topological reasoning, place-cell encoding, and episodic memory to guide behaviour. The agent builds and updates a sparse topological map online, learns state transitions dynamically, and plans actions by minimising Expected Free Energy. This allows it to balance goal-directed and exploratory behaviours. We implemented a ROS-compatible navigation system that is sensor and robot-agnostic, capable of integrating with diverse hardware configurations. It operates in a fully self-supervised manner, is resilient to drift, and supports both exploration and goal-directed navigation without any pre-training. We demonstrate robust performance in large-scale real and simulated environments against state-of-the-art planning models, highlighting the system's adaptability to ambiguous observations, environmental changes, and sensor noise. The model offers a biologically inspired, modular solution to scalable, self-supervised navigation in unstructured settings. AIMAPP is available at https://github.com/decide-ugent/AIMAPP.

Problem

• Autonomous robot navigation without predefined maps or training
• Simultaneous exploration, localization and planning under uncertainty
• Balancing goal-directed and exploratory behaviors in unfamiliar environments

Innovation

• Active Inference unifies mapping, localisation, and decision-making
• Topological reasoning with place-cell encoding guides behaviour
• ROS-compatible system operates fully self-supervised without pre-training