CABLE: Cloud-Assisted Bandwidth-efficient LMM-based Encoding for V2X Systems

2026-06-17

AI Summary

This work addresses the high communication overhead and substantial prefill latency of large multimodal models (LMMs) in cloud-based V2X systems caused by uploading full-resolution video streams. To mitigate these issues, the authors propose a cloud-assisted, bandwidth-efficient encoding framework wherein edge nodes propagate segmentation masks from the cloud using ego-motion compensation and dynamically generate regions of interest (ROIs) by integrating residual motion cues with corridor envelopes. Only ROI patches are transmitted to the cloud, and a mask-guided closed-loop feedback mechanism is established to maintain semantic consistency. Evaluated across five datasets, the approach reduces ROI pixel transmission by 73%–87% and accelerates LMM prefilling by 5–8×, while incurring only minor and controllable degradation in perception performance.

Abstract

Cloud-hosted large multimodal models (LMMs) can provide strong open-vocabulary perception for Vehicle-to-Everything systems, but naively transmitting full-resolution frames from edge to cloud causes severe communication overhead and high cloud-side prefill latency. We present CABLE, a cloud-assisted bandwidth-efficient LMM-based encoding framework for edge-cloud perception. CABLE propagates the previous cloud segmentation mask on the edge using ego-motion compensation, refines it with residual-motion cues, and consolidates disconnected regions via a corridor envelope to form a robust region of interest (ROI). Only ROI-masked images are uploaded, while the cloud segmentation output is fed back as the prior for the next frame, forming a mask-to-ROI-to-LMM feedback loop. Experiments on five datasets (nuScenes, WOD-ZB, Waymo, KITTI, and CADC) show consistent communication savings while largely preserving perception, achieving $73$--$87\%$ ROI pixel-coverage reduction with $5$--$8\times$ estimated LMM prefill speedup at a modest detection-quality trade-off relative to full-frame inference.

Problem

• V2X
• large multimodal models
• communication overhead
• cloud-edge perception
• prefill latency

Innovation

• bandwidth-efficient encoding
• cloud-assisted perception
• region of interest (ROI)
• large multimodal models (LMMs)
• edge-cloud feedback loop