Software-Defined 5G/6G IoV Networks

• Software-Defined 5G/6G IoV networks are next-generation vehicular architectures that integrate SDN, NFV, and dual PID controllers for dynamic load and energy management.
• The architecture employs a three-layer framework where SDN controllers, embedded with PID logic, orchestrate network flows and resource allocation for scalable, real-time communications.
• Experimental outcomes demonstrate up to 30% energy savings, improved load balancing, and enhanced QoS/QoE, validating the framework's effectiveness in smart city environments.

Software-defined 5G/6G Internet of Vehicles (IoV) networks represent an architectural paradigm where Software-Defined Networking (SDN) and Network Functions Virtualization (NFV) are tightly integrated to enable elastic, centralized, and energy-efficient networking for connected vehicles in smart city contexts. The proliferation of multimedia-driven vehicular applications over emerging 5G and 6G infrastructure imposes stringent demands on scalability, adaptivity, and especially resource and energy management. A recent approach leverages dual Proportional-Integral-Derivative (PID) controllers to holistically regulate both resource load-distribution and system temperature, providing dynamic, self-adaptive orchestration across the IoV stack (Montazerolghaem, 1 Feb 2026).

1. Layered Framework Architecture

The core design is a three-layer end-to-end architecture combining SDN and NFV, structuring the IoV network into distinct functional strata:

• Application Plane: Houses the Energy Management and Load-Balancing applications, both incorporating PID logic. This layer sets global policies, including target CPU utilization, temperature thresholds, and migration strategies.
• Control Plane: Implements the SDN controller, using software such as RYU, NOX, or POX, interfacing with all OpenFlow switches. This plane features two logical sub-controllers:
  • Flow Controller (PID-embedded): Assigns per-flow processing to physical machines (PMs) according to current and forecasted load.
  • VNF Controller (PID-embedded): Manages PM power states (on/off), VNF placement, and live migrations.
  • Key modules include Network Statistics, PM Workload Monitoring, Workload Estimation via NLMS, PM Resource and Temperature Monitoring, and modular PID Engines, with Flow and VNF Managers executing the mobility and orchestration logic.
• Data Plane: Comprises OpenFlow switches linking Road-Side Units (RSUs; 5G/6G gNB/eNB) and commodity servers (PMs). Servers host multimedia Virtual Network Functions (VNFs) within VMs, forming the NFV Infrastructure (NFVI). Vehicular mobility is emulated by SUMO, with Mininet and OpenStack creating a realistic virtualization and switching environment (Montazerolghaem, 1 Feb 2026).

2. Dual PID Controller Methodology

Two discrete-time PID controllers operate concurrently within the SDN controller, targeting orthogonal aspects of system dynamics:

• Load-Distribution PID: Controls aggregate CPU utilization by dynamically adjusting per-PM load thresholds. The PID law,

$$u(t+1) = K_p^L e_L(t+1) + K_i^L \sum_{\tau=0}^{t+1} e_L(\tau) \Delta t + K_d^L \frac{e_L(t+1)-e_L(t)}{\Delta t},$$

is used, where $e_L(t+1)$ is the load error (predicted minus target load). Once $u(t+1)$ is computed and clamped, the controller selects routing rules—either “Least-Connections” or “Least-Response-Time”—to optimize allocation (Montazerolghaem, 1 Feb 2026).

• Temperature-Regulation PID: Manages thermal load by activating or deactivating PMs and migrating VNFs. The PID controller

$$w(t+1) = K_p^T e_T(t+1) + K_i^T \sum_{\tau=0}^{t+1} e_T(\tau) \Delta t + K_d^T \frac{e_T(t+1)-e_T(t)}{\Delta t}$$

uses temperature error $e_T(t+1)$. Decision logic determines whether to power on additional PMs and rebalance VNFs, or consolidate VNF instances and power off underutilized servers.

Control actions in both subsystems are bounded (thresholded) to avoid oscillations and resource "thrashing," ensuring system stability despite highly variable vehicular and multimedia loads (Montazerolghaem, 1 Feb 2026).

3. Resource and Energy Modeling

Resource prediction and energy management is achieved through a combination of statistical filtering and threshold-based mapping:

• CPU Utilization Prediction: Each PM samples local CPU load at regular intervals ($\Delta t$), storing windowed samples $\mathbf{X}_i(t)$. A normalized least-mean-square (NLMS) filter recursively predicts $\hat{x}_i(t+1)$, which feeds into the global load predictor $V(t+1)$.
• Temperature Estimation: CPU predictions are mapped to normalized temperature using a piecewise linear function, defined by thresholds $A$ and $B$, enabling control-law-based power-state transitions.
• Energy Consumption Measurement: Each PM logs real-time power draw, with total energy computed as $\sum (\text{PM}_i~\text{power} \times \Delta t)$. Power savings are evaluated relative to a static, non-SDN/NFV baseline.

This modeling underpins the control loop’s proactive, predictive balancing of service quality and energy use across the distributed infrastructure (Montazerolghaem, 1 Feb 2026).

4. Control Loop and Algorithms

Network operation comprises a periodic control loop with distinct actionable steps:

1. Monitoring: FlowManager and NetworkStats gather real-time telemetry from OpenFlow switches and PMs.
2. Prediction: PM CPU time series are filtered and extrapolated via NLMS, producing next-step load and temperature forecasts.
3. Control Computation: Load and temperature errors ($e_L,e_T$) are calculated against set-points; PID controllers update their respective actuators ($u,w$).
4. Actuation:
   • Flow distribution is selected based on $u$ (connections or response time metric).
   • VNF placement and PM state transitions executed according to $w$ (temperature response).
   • Specific algorithms (TemperatureIncrease/Reduction) consolidate or expand active PMs and redistribute VNFs accordingly.
5. Iteration: Loop repeats for each sampling interval (Montazerolghaem, 1 Feb 2026).

5. Experimental Setup and Key Metrics

Simulation topology incorporates:

• Emulation Tools: Mininet + OpenStack for switching and virtualization, RYU/NOX SDN controller, SUMO for urban vehicular mobility.
• Testbed: 3 RSUs, 8 OpenFlow switches, 3 PMs, and 22 wired links; multimedia streaming workloads for 5–100 vehicles.
• Baselines: Traditional (static allocation, no SDN/NFV), and SDN-only (no PID logic).
• KPIs:
  • Load balancing (standard deviation of per-PM CPU)
  • Mean CPU utilization (PMs/switches)
  • Total energy consumption (Wh)
  • Throughput, request response rate (%)
  • QoS indicators (delay, setup time, jitter)
  • QoE metrics (MOS, R-Factor, packet-loss rate)

6. Performance Outcomes and Analysis

Under intensive traffic (≈1500 requests/s), the PID-controlled software-defined framework yields:

• Energy Efficiency: Total power consumption reduced by ≈30% (peak PM power: 300 W→210 W) compared to static baseline.
• Load Balancing: Average per-PM CPU stabilized at ≈33%±2%, compared to skewed baseline allocation (70%/20%/10% split).
• Scalability and Responsiveness: 98%+ request response rates at peak, controller CPU usage capped under 70%, with memory well below system thresholds.
• QoS/QoE Gains: Lower delay (15 ms vs. 25 ms), faster setup, reduced jitter, improved MOS (3.5→4.6), R-Factor (65→85), and marked reductions in packet loss.
• Why Effective: Proactive NLMS-PID prediction, centralized SDN-based flow allocation, and elastic NFV orchestration minimize both idle-server power waste and overloads, directly addressing dynamic multimedia traffic volatility (Montazerolghaem, 1 Feb 2026).

7. Challenges and Future Research Directions

Limitations identified include exclusive CPU-centric modeling (no explicit memory, storage, or bandwidth control), assumptions around rapid VNF live-migration (<100 ms), and unaddressed issues in multi-tenant security and 6G RAN slicing. Emerging research priorities encompass:

• Extension to multi-resource PID regulation
• AI-based auto-tuning of PID gains
• Integration with Multi-Access Edge Computing (MEC) for ultra-low latency targeting sub-millisecond response
• Validation in real-world vehicular deployments under high mobility

These directions are expected to drive further optimization and adaptation of software-defined 5G/6G IoV networks for diverse and evolving smart city environments (Montazerolghaem, 1 Feb 2026).