Digital twin-assisted multi-layer network: resource optimization for low-latency and energy-efficiency
Abstract
The sixth-generation (6G) wireless networks are expected to provide ubiquitous connectivity, high
data rate, low latency, energy efficiency, and edge intelligence for Internet of Things (IoT) applications. However, effective resource scheduling and network configuration in 6G is challenging
due to the resource-constrained devices, high quality-of-service (QoS) requirement, and high density of heterogeneous devices. Multi-layer networks are potential candidates for addressing the
challenge of resource-constrained devices to meet their tasks’ QoS requirements. Still, there is the
challenge of resource scheduling and management of multi-layer networks. Digital twin technology is a promising solution to enable multi-layer wireless networks that incorporate IoT devices on
the ground, unmanned aerial vehicles (UAVs) as mobile edge computing (MEC) servers, and cloud
servers. Multi-layer processing can handle time-sensitive and computationally intensive tasks from
IoT devices. In this thesis, we propose a digital twin-assisted multi-layer network for low-latency
and energy-efficient communication and computation. We mathematically formulate an optimization problem to minimize task latency and energy consumption of IoT devices by optimizing their
association with the UAV-MECs, computation resources, communication resources, and offloading
portions of tasks. The formulated problem is a non-linear and non-convex optimization problem.
We propose a two-stage scheme based on the K-means method and the deep neural network approach to solve the above optimization problem. The K-means method is utilized for the optimal
placement of UAV-MECs in the first stage, and then we associate the IoT devices with UAV-MECs
for offloading tasks. In the second stage, the deep learning architecture is utilized to optimize network resources. We compare the proposed two-stage scheme with existing schemes to highlight
the scalability of the proposed solution. We perform extensive simulations by varying the number
of UAV-MECs and IoT devices in the network to look at the impact on task latency and energy
consumption by IoT devices. Fixed offloading portioning is compared with optimized offloading
portioning to highlight the usefulness of optimization in terms of latency and energy minimization.
Simulation results demonstrate the usefulness of the multi-layer network in achieving low latency
and energy-efficient computation and communication.