AdaSprite: Resource-efficient Online Co-Adaptation for V2I Systems Under Large-scale Data Drifts

2026-08-17

AI Summary

This study addresses the low adaptive efficiency of resource-constrained edge V-MoE models under data drift in Vehicle-to-Infrastructure (V2I) scenarios by proposing AdaSprite. The system integrates collaborative elastic scaling, multi-level multiplexing, and double-buffer scheduling to effectively mitigate memory fragmentation and I/O bottlenecks, thereby surpassing concurrent task limitations. Experimental results demonstrate that AdaSprite supports 17 concurrent tasks while improving SLO attainment by 1.6× and throughput by 2.1×. Furthermore, it achieves second-level online adaptation, significantly enhancing both multi-task processing capabilities and resource utilization efficiency at the edge.

Abstract

The rise of vehicle-infrastructure (V2I) collaboration enables safer and broader perception. To process large-scale V2I video streams, vision-language models (VLMs) are promising as they unify multi-view vision into end-to-end task grounding, reducing handcrafted design. We use Vision Mixture-of-Experts (V-MoE) as the distributed visual backbone of VLMs, leveraging sparse expert routing to enable conditional computation across diverse viewpoints under resource constraints. Yet, V-MoEs face a critical challenge: large-scale data shifts over minutes to hours in V2I systems, amplified by agnostic participants and biased features propagating through experts. To maintain accuracy efficiently, we find it beneficial to co-adapt multiple V-MoEs on edge servers, avoiding the latency and privacy risks of cloud offloading and the accuracy sacrifices of on-device methods. However, the resource-constrained edge poses challenges for efficient co-adaptation: i) DRAM fragmentation and imbalance limit expert parallelism, ii) memory-I/O bottlenecks restrict computation reuse, and iii) asynchronous adaptation increases task-switch overhead. Also, prior work rarely explores the upper bound of concurrent tasks under limited edge resources, a critical factor for practical V2I deployment. To address these, we present AdaSprite. By combining cooperative elastic scaling with multi-level multiplexing, AdaSprite optimizes expert lifespans to reduce DRAM fragmentation, exploits predictable activation patterns for efficient I/O reuse, and employs twin-buffer scheduling to leverage sparsity. On a weak edge, AdaSprite supports up to 17 concurrent V2I tasks (vs. up to 6 for baselines), improving SLO attainment by 1.6x and throughput by 2.1x. Also, it allows users to trade accuracy and concurrency for second-level adaptation.

Problem

• Vehicle-Infrastructure Collaboration
• Vision Mixture-of-Experts
• Online Co-Adaptation
• Edge Computing
• Data Drift

Innovation

• Vision Mixture-of-Experts
• Edge Co-Adaptation
• Elastic Scaling
• Twin-buffer Scheduling
• V2I Systems