![[2602.15836] EdgeNav-QE: QLoRA Quantization and Dynamic Early Exit for LAM-based Navigation on Edge Devices](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.15836] EdgeNav-QE: QLoRA Quantization and Dynamic Early Exit for LAM-based Navigation on Edge Devices
Summary
The paper presents EdgeNav-QE, a framework that combines QLoRA quantization and dynamic early exit mechanisms to enhance LAM-based navigation on edge devices, significantly reducing latency and memory usage while maintaining high navigation success rates.
Why It Matters
As autonomous navigation systems become increasingly vital in robotics, optimizing their performance for edge devices is crucial. This research addresses the challenges of deploying large models in resource-constrained environments, making it relevant for advancements in robotics and AI applications.
Key Takeaways
- EdgeNav-QE integrates QLoRA quantization with dynamic early-exit for efficient navigation.
- The framework reduces inference latency by 82.7% and memory footprint by 66.7%.
- It maintains an 81.8% navigation success rate, outperforming static early-exit methods.
- The approach is particularly beneficial for safety-critical applications requiring real-time processing.
- The study demonstrates the effectiveness of content-aware adaptive computation in edge scenarios.
Computer Science > Robotics arXiv:2602.15836 (cs) [Submitted on 12 Jan 2026] Title:EdgeNav-QE: QLoRA Quantization and Dynamic Early Exit for LAM-based Navigation on Edge Devices Authors:Mengyun Liu, Shanshan Huang, Jianan Jiang View a PDF of the paper titled EdgeNav-QE: QLoRA Quantization and Dynamic Early Exit for LAM-based Navigation on Edge Devices, by Mengyun Liu and 2 other authors View PDF HTML (experimental) Abstract:Large Action Models (LAMs) have shown immense potential in autonomous navigation by bridging high-level reasoning with low-level control. However, deploying these multi-billion parameter models on edge devices remains a significant challenge due to memory constraints and latency requirements. In this paper, we propose EdgeNav-QE, a novel framework that integrates Quantized Low-Rank Adaptation (QLoRA) with a dynamic early-exit (DEE) mechanism to optimize LAMs for real-time edge navigation. By quantizing the backbone to 4-bit precision and strategically placing early-exit branches, we enable the model to terminate inference early for simple navigation tasks while retaining full depth for complex decision-making. Experimental results on the Habitat-Sim environment with Matterport3D dataset using OpenVLA-7B backbone, demonstrate that EdgeNav-QE reduces inference latency by 82.7% and memory footprint by 66.7% compared to full-precision baselines, while maintaining 81.8% navigation success rate. Furthermore, it outperforms state-of-the-art static early-exit m...
