AdaSprite：大规模数据漂移下车路协同（V2I）系统的资源高效在线协同自适应
AdaSprite: Resource-efficient Online Co-Adaptation for V2I Systems Under Large-scale Data Drifts
AI总结：
针对V2I系统大规模数据漂移下边缘资源受限的协同自适应问题，AdaSprite通过协同弹性扩缩等技术，使弱边缘支持17个并发V2I任务，SLO达成率与吞吐量分别提升1.6倍、2.1倍。
AI中文摘要：
车路协同（V2I）的兴起可实现更安全、更广泛的感知。为处理大规模V2I视频流，视觉-语言模型（VLMs）极具应用前景，因其可将多视角视觉统一为端到端任务定位，减少手工设计。我们采用视觉混合专家模型（V-MoE）作为VLMs的分布式视觉骨干，利用稀疏专家路由在资源约束下实现不同视角的条件计算。然而，V-MoE面临一个关键挑战：V2I系统中每几分钟至数小时就会出现大规模数据漂移，且该漂移因不可知的参与者及通过专家传播的偏置特征而加剧。为高效维持准确率，我们发现于边缘服务器上协同自适应多个V-MoE是有益的，可避免云端卸载带来的延迟与隐私风险，以及设备端方法的准确率牺牲。但资源受限的边缘对高效协同自适应构成挑战：i）DRAM碎片化与失衡限制专家并行性；ii）内存-I/O瓶颈限制计算复用；iii）异步自适应增加任务切换开销。此外，现有研究极少探讨有限边缘资源下并发任务的上限，这是V2I实际部署的关键因素。为解决这些问题，我们提出AdaSprite。通过结合协同弹性扩缩与多级复用，AdaSprite优化专家生命周期以减少DRAM碎片化，利用可预测的激活模式实现高效I/O复用，并采用双缓冲调度以利用稀疏性。在弱边缘设备上，AdaSprite支持多达17个并发V2I任务，而基线方法仅支持最多6个，服务水平目标（SLO）达成率提升1.6倍，吞吐量提升2.1倍。此外，它允许用户在秒级自适应范围内权衡准确率与并发度。
英文摘要：
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.
