Robust service migration for autonomous vehicles leveraging deep learning and cooperative V2V protocols.

Ahmadpanah, SH; Sahafi, A; Erfani, SH · 2026 · PubMed Central

DOI: 10.1038/s41598-026-48717-7

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of maintaining service continuity for autonomous vehicles during high-speed mobility, where frequent handovers between Mobile Edge Computing (MEC) nodes are required. Current reactive migration methods, such as Pre-copy and Post-copy, suffer from excessive downtime and bandwidth waste because they fail to predict memory modifications or account for data semantics. The authors propose a "Cognitive Seamless Service Migration" framework that leverages deep learning to proactively optimize the transfer of containerized services, aiming to reduce latency and packet loss in volatile vehicular networks. The proposed system integrates five core components. First, a Deep Temporal Predictive Filtering mechanism uses Long Short-Term Memory (LSTM) networks to analyze historical memory traces and predict the probability of future page modifications, allowing the system to skip pages likely to become dirty. Second, a Content-Aware Semantic Compression engine classifies data types (e.g., sensor logs, LiDAR, code) using a shallow CNN and applies tailored compression strategies, such as hybrid autoencoders for high-entropy data and Zstandard for executable code. Third, a Deep Reinforcement Learning (DRL) agent, trained with Proximal Policy Optimization, dynamically determines the optimal timing for the final "Stop-and-Copy" phase by monitoring network bandwidth, signal-to-noise ratio, and vehicle velocity. Fourth, a Lightweight Merkle Tree scheme enables granular, chunk-level integrity verification to avoid full-file hashing delays. Finally, a Multi-Path Cooperative Migration strategy utilizes Vehicle-to-Vehicle (V2V) communication to offload non-critical data to neighboring vehicles when infrastructure links are congested. Experimental simulations using realistic urban mobility models demonstrate that this cognitive approach significantly outperforms standard adaptive methods. The framework reduced wasted data transmission by 28% and slashed service downtime by 34%. The LSTM-based prediction effectively minimized redundant transfers of volatile memory pages, while the semantic compression improved bandwidth efficiency by matching algorithms to data characteristics. The DRL agent successfully navigated stochastic network fluctuations, initiating handovers at optimal moments to avoid link failures. The cooperative V2V protocol further enhanced robustness by aggregating bandwidth during congestion events. The significance of this work lies in its shift from reactive to predictive service migration in the Internet of Vehicles. By integrating temporal prediction, semantic awareness, and reinforcement learning, the framework provides a robust solution for maintaining strict Service Level Agreements in high-mobility scenarios. This approach ensures stable digital handovers even under challenging physical conditions, offering a scalable model for future autonomous vehicle infrastructure where low-latency processing is critical.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• v2x connected vehicle