Risk-Adaptive Edge--Cloud Visual Reasoning for Communication-Efficient Autonomous Driving

2026-08-14

AI Summary

This study addresses the high communication overhead and latency bottlenecks associated with cloud-based vision-language model inference for autonomous driving by proposing a risk-adaptive edge-cloud collaborative architecture. The framework incorporates a selective cloud offloading mechanism driven by local risk assessment, utilizing a lightweight detector to dynamically trigger cloud inference and achieve an adaptive balance between communication efficiency and safety. Experimental results demonstrate that this architecture reduces cloud request volume by 54.1% while maintaining task success rates and decreasing emergency braking frequency. Furthermore, the system effectively mitigates traffic load under network latency conditions, significantly enhancing overall robustness in autonomous driving applications.

Abstract

Cloud-hosted vision-language models (VLMs) offer greater contextual reasoning capabilities than smaller onboard models, but frequent visual uploads increase communication overhead and add network and inference latency to tactical decisions. We present a risk-adaptive edge-cloud architecture in which onboard traffic assessment determines when cloud reasoning is requested. An onboard VLM and a lightweight detector capture temporal traffic conditions and path-relative hazards for conservative local response and selective cloud access. The cloud model provides tactical advice, while validation, vehicle control, and automatic emergency braking remain local. In CARLA experiments, our method matched the task success rate of periodic cloud access while reducing cloud requests by 54.1% and recording fewer automatic emergency braking (AEB) activations. In a delayed-roadwork ablation, semantic events triggered requests before the next scheduled audit. Across three emulated network profiles, the method continued to reduce cloud traffic, although lane changes took longer than with periodic access. Onboard traffic assessment therefore served as a practical trigger for selective VLM inference in these experiments.

Problem

• Autonomous Driving
• Vision-Language Models
• Communication Overhead
• Edge-Cloud Collaboration
• Inference Latency

Innovation

• Risk-Adaptive Edge-Cloud Architecture
• Selective Cloud Inference
• Vision-Language Models
• Communication-Efficient Autonomous Driving
• Onboard Traffic Assessment