Towards Lifelong Aerial Autonomy: Geometric Memory Management for Continual Visual Place Recognition in Dynamic Environments

2026-04-10

AI Summary

This work addresses catastrophic forgetting in visual place recognition under long-term aerial autonomous missions, where dynamic environmental changes pose significant challenges. The problem is formulated as a domain-incremental learning task, and a heterogeneous memory framework is proposed to operate under strict onboard storage constraints. The framework employs a “learn-and-discard” mechanism that decouples static satellite-derived anchor points from a dynamic experience replay buffer, complemented by a spatial constraint strategy to optimize sample selection. Innovatively integrating static geometric priors with dynamic feature replay, the approach prioritizes diversity in memory structure over sample difficulty, thereby substantially improving the trade-off between plasticity and stability as well as sequence-agnostic robustness. Evaluated on a benchmark of 21 sequential tasks, the method achieves a 7.8% improvement in knowledge retention over random buffer selection and demonstrates enhanced spatial generalization capability.

Technology Category

• Computer Vision: Representation Learning for Vision
• Search and Optimization: Learning to Search
• Machine Learning: Life-Long and Continual Learning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Robust geo-localization in changing environmental conditions is critical for long-term aerial autonomy. While visual place recognition (VPR) models perform well when airborne views match the training domain, adapting them to shifting distributions during sequential missions triggers catastrophic forgetting. Existing continual learning (CL) methods often fail here because geographic features exhibit severe intra-class variations. In this work, we formulate aerial VPR as a mission-based domain-incremental learning (DIL) problem and propose a novel heterogeneous memory framework. To respect strict onboard storage constraints, our"Learn-and-Dispose"pipeline decouples geographic knowledge into static satellite anchors (preserving global geometric priors) and a dynamic experience replay buffer (retaining domain-specific features). We introduce a spatially-constrained allocation strategy that optimizes buffer selection based on sample difficulty or feature space diversity. To facilitate systematic assessment, we provide three evaluation criteria and a comprehensive benchmark derived from 21 diverse mission sequences. Extensive experiments demonstrate that our architecture significantly boosts spatial generalization; our diversity-driven buffer selection outperforms the random baseline by 7.8% in knowledge retention. Unlike class-mean preservation methods that fail in unstructured environments, maximizing structural diversity achieves a superior plasticity-stability balance and ensures order-agnostic robustness across randomized sequences. These results prove that maintaining structural feature coverage is more critical than sample difficulty for resolving catastrophic forgetting in lifelong aerial autonomy.

Problem

• lifelong aerial autonomy
• visual place recognition
• catastrophic forgetting
• domain-incremental learning
• dynamic environments

Innovation

• continual learning
• visual place recognition
• domain-incremental learning
• experience replay
• geometric memory