EdgeLPR: On the Deep Neural Network trade-off between Precision and Performance in LiDAR Place Recognition

2026-05-04

AI Summary

Deploying high-accuracy LiDAR place recognition models on resource-constrained edge platforms entails a trade-off between computational efficiency and recognition performance. This work addresses this challenge by transforming LiDAR data into bird’s-eye-view representations and formulating place recognition as a lightweight image classification task, employing a unified descriptor architecture based on global pooling followed by linear projection. To guide practical deployment, the study introduces a use-case-aware quantization evaluation framework tailored for edge AI, systematically assessing multiple lightweight networks under FP32, FP16, and INT8 precision settings. Experimental results demonstrate that FP16 quantization achieves substantial computational savings with negligible accuracy loss, whereas the performance degradation under INT8 is highly architecture-dependent, offering actionable insights and design principles for efficient edge deployment of LiDAR-based place recognition systems.

Abstract

Place recognition is essential for long-term autonomous navigation, enabling loop closure and consistent mapping. Although deep learning has improved performance, deploying such models on resource-constrained platforms remains challenging. This work explores efficient LiDAR-based place recognition for EdgeAI by leveraging Bird's Eye View representations to enable lightweight image-based networks. We benchmark representative architectures without aggregation heads using a unified descriptor scheme based on global pooling and linear projection, and evaluate performance under FP32, FP16, and INT8 quantization. Experiments reveal trade-offs between accuracy, robustness, and efficiency: FP16 matches FP32 with lower cost, while INT8 introduces architecture-dependent degradation. Overall, the presented results are a strong basis for future research on 'use-case'-aware quantisation of Neural Networks for Edge deployment.

Problem

• LiDAR Place Recognition
• EdgeAI
• Precision-Performance Trade-off
• Quantization
• Resource-constrained Deployment

Innovation

• EdgeAI
• LiDAR Place Recognition
• Bird's Eye View
• Neural Network Quantization
• Efficient Deep Learning