Multi-Layer Latency Aware Workload Assignment of E-Transport IoT Applications in Mobile Sensors Cloudlet Cloud Networks

Lakhan, Abdullah; Dootio, Mazhar Ali; Groenli, Tor Morten; Sodhro, Ali Hassan; Khokhar, Muhammad Saddam · 2021 · OpenAlex-citations

DOI: 10.3390/electronics10141719

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 minimizing latency for latency-sensitive E-Transport IoT applications, such as e-buses, e-taxis, and autonomous vehicles, within Mobile Sensors Cloudlet Cloud (MCBC) networks. Conventional cloud computing often incurs high end-to-end latency due to distance, while existing cloudlet-based solutions frequently ignore critical delay factors like round-trip and migration delays, or fail to adapt to the dynamic mobility of E-Transports. The authors propose a novel Multi-Layer Latency Aware Workload Assignment Strategy (MLAWAS) designed to allocate workloads to optimal computing nodes by accounting for communication, round-trip, process, and migration delays. The primary objective is to minimize the average response time of applications while maintaining Quality of Service (QoS) requirements in a dynamic environment prone to overloading and overheating. The study employs a hybrid algorithmic framework combining global and local search techniques with reinforcement learning. The system model utilizes an M/M/1 queuing system to characterize workload arrivals via a Poisson process and calculates processing and transmission delays. The MLAWAS framework consists of four key components: an Improved Genetic Algorithm (IGA) for global optimization, combined with Simulated Annealing to enhance probabilistic hill-climbing capabilities; a local greedy search to verify optimal average response times for assigned workloads; and a Q-Learning aware migration technique. This Q-Learning component treats the current optimal solution as a state, allowing the system to dynamically migrate Virtual Machines (VMs) in response to environmental changes, such as E-Transport mobility or node overload, thereby preventing resource constraints and maintaining load balance. Simulation results demonstrate that MLAWAS achieves the minimum average response time compared to two existing workload assignment strategies. By integrating multi-layer latency considerations—including often-ignored migration and round-trip delays—into the assignment logic, the proposed method effectively handles the coarse-grained offloading of E-Transport workloads. The Q-Learning mechanism specifically ensures that the system adapts to dynamic changes, such as the movement of vehicles and the resulting need for workload migration between cloudlets and remote clouds, without incurring excessive overhead. The significance of this work lies in its comprehensive approach to latency optimization in distributed edge computing environments. Unlike previous studies that relied on static assumptions or ignored migration costs, MLAWAS provides a robust solution for dynamic, mobility-aware networks. It offers a guaranteed global optimal solution for workload assignment, addressing the limitations of greedy and heuristic-only approaches that struggle in unstable environments. This contributes to the development of more efficient, low-latency infrastructure for critical E-Transport applications, ensuring reliable performance despite the complexities of mobile sensor networks and heterogeneous cloudlet resources.

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.

• teleoperation remote driving