Ranking-aware Continual Learning for LiDAR Place Recognition

2025-05-12

AI Summary

To address catastrophic forgetting in LiDAR place recognition (LPR) under continual learning, this paper proposes a Knowledge Distillation and Fusion (KDF) framework. Methodologically, it introduces (1) a novel ranking-aware knowledge distillation loss that explicitly embeds retrieval ranking structure into the distillation objective, thereby preserving high-level semantic discriminative knowledge from previous tasks; and (2) a lightweight knowledge fusion module that enables collaborative optimization of feature representations between old and new models. Evaluated on multiple LiDAR benchmarks, the approach significantly mitigates forgetting: it achieves an average Recall@1 surpassing current state-of-the-art methods and reduces the forgetting score by 12.7%. The framework offers a scalable, low-overhead solution for continual deployment of LiDAR-based scene recognition systems.

Technology Category

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

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Place recognition plays a significant role in SLAM, robot navigation, and autonomous driving applications. Benefiting from deep learning, the performance of LiDAR place recognition (LPR) has been greatly improved. However, many existing learning-based LPR methods suffer from catastrophic forgetting, which severely harms the performance of LPR on previously trained places after training on a new environment. In this paper, we introduce a continual learning framework for LPR via Knowledge Distillation and Fusion (KDF) to alleviate forgetting. Inspired by the ranking process of place recognition retrieval, we present a ranking-aware knowledge distillation loss that encourages the network to preserve the high-level place recognition knowledge. We also introduce a knowledge fusion module to integrate the knowledge of old and new models for LiDAR place recognition. Our extensive experiments demonstrate that KDF can be applied to different networks to overcome catastrophic forgetting, surpassing the state-of-the-art methods in terms of mean Recall@1 and forgetting score.

Problem

• Addresses catastrophic forgetting in LiDAR place recognition
• Proposes ranking-aware knowledge distillation for preserving recognition knowledge
• Introduces knowledge fusion to integrate old and new model knowledge

Innovation

• Ranking-aware knowledge distillation loss
• Knowledge fusion module integration
• Overcoming catastrophic forgetting effectively