Abstract
Smart Manufacturing has become a popular topic due to the growth of Internet of Things (IoT) and Cloud Manufacturing, which is driving traditional manufacturing companies to migrate towards digitization and intelligence. Device to Device (D2D) facilitates communication between terminal devices and improves output and operational efficiency. Firstly, by using the utility theory of microeconomics, this paper constructs a novel utility-oriented resource allocation model and addresses the resource allocation problem for D2D communication. This paper also proposes an interesting price-based resource allocation scheme along with an improved gradient descent method and proves the stability of the price-based resource allocation technique based on the Lyapunov stability theory. The simulation results demonstrate that the two proposed algorithms can achieve the equilibrium which is also the optimum of the resource allocation model, and that the gradient descent method is outperformed by the price-based mechanism method in terms of stability and convergence speed.
Introduction
In the era of Smart Manufacturing and the Industrial Internet of Things (IIoT), factories rely on a massive web of sensors, robots, and high-definition cameras. Traditionally, these devices communicate via a central cloud. However, "long-haul" cloud communication introduces significant latency, which is unacceptable for real-time tasks like a robot arm reacting to a sensor or a machine vision system detecting defects on a high-speed conveyor belt.
Device-to-Device (D2D) communication solves this by allowing terminal devices to talk directly to one another. The challenge lies in resource allocation: how to fairly and efficiently distribute limited bandwidth among dozens of competing devices without a central "brain" micromanaging every packet.
Methodology
The authors tackle this by modeling the factory network as a microeconomic ecosystem where bandwidth is treated as a market commodity.
Utility-Oriented Modeling
They define a "utility function" for each device using a logarithmic form ($U_s = \omega_s \ln y_s$). This ensures proportional fairness, meaning no single device can hog all the bandwidth while others starve.
The Price-based Mechanism
The core contribution is a decentralized approach where "prices" (Lagrange multipliers) are assigned to bandwidth.
- Demanders (e.g., monitors) calculate how much they are willing to pay based on their needs.
- Providers (e.g., HD cameras or UAVs) update their prices based on total demand and their own capacity.
Improved Gradient Descent
As a baseline, the authors also implemented a gradient descent method enhanced with low-pass filtering. This helps eliminate the oscillations and fluctuations typically found in non-strictly concave optimization problems. The authors provide a rigorous mathematical proof using Lyapunov stability theory to guarantee that their price-based system will always converge to a stable, optimal equilibrium.
Key Insights and Results
The most significant insight is that decentralized "pricing" naturally leads to global optimization. When the price of a provider's bandwidth rises due to high demand, devices naturally shift their "purchasing" to less congested providers.
Simulation and Performance
The authors designed a numerical simulation representing a typical smart manufacturing environment with 8 devices (2 resource providers and 6 demanders).
- Convergence to Optimum: Both the improved gradient descent and the price-based mechanism successfully reached the global optimum (verified by LINGO), achieving a total network utility of approximately 365.64.
- Speed and Stability: The price-based mechanism significantly outperformed gradient descent. It reached equilibrium in fewer iterations and showed much smoother convergence, whereas the gradient descent method experienced a period of fluctuation before settling.
Conclusion
This paper introduces a decentralized, price-based resource allocation scheme for D2D communication in smart factories. By treating bandwidth as a market commodity and applying microeconomic utility theory, the proposed algorithm ensures low-latency data exchange for critical industrial tasks. While the current model assumes static scenarios, it provides a strong foundation for future research into mobile IoT devices and real-world interference in smart manufacturing environments.
Summary updated