Advanced International Journal for Research (AIJFR)

E-ISSN: 3048-7641 ● Website: [Link] ● Email: editor@[Link]

A Reinforcement Learning-Based Adaptive

Routing Framework for Real-Time Optimization
in SD-WAN Environments
Rahul Guha1, Nidhi Singh2, Ambika Bagri3, Pankaj Kumar Sharma4
1,2,3,4
Assistant Professor, Artificial Intelligence & Data Science, JECRC Foundation (RTU, Kota)

Abstract

The rapid evolution of enterprise networks and the growing demand for application-specific Quality of
Service (QoS) have made Software-Defined Wide Area Networks (SD-WANs) a crucial component of
modern communication infrastructures. However, traditional static or rule-based routing approaches often
fail to adapt efficiently to dynamic network conditions such as fluctuating link performance, congestion,
and heterogeneous transport mediums. This paper proposes a Reinforcement Learning-Based Adaptive
Routing Framework (RLARF) designed to optimize routing decisions in real-time for SD-WAN
environments. The framework employs a deep reinforcement learning agent that continuously learns from
network states, including latency, packet loss, and bandwidth utilization, to select optimal routing paths
dynamically. By formulating the routing process as a Markov Decision Process (MDP), the proposed
model efficiently adapts to varying link qualities and application QoS requirements while ensuring
scalability across large, distributed networks. Experimental evaluations demonstrate that RLARF achieves
significant improvements in throughput, latency reduction, and reliability compared to traditional routing
algorithms. Furthermore, the proposed system exhibits robust performance under highly variable network
conditions, making it a viable solution for next-generation, intelligent SD-WAN architectures.

Keywords: Reinforcement Learning, SD-WAN Optimization, Adaptive Routing

1. Introduction

In recent years, the exponential growth of cloud computing, Internet of Things (IoT) applications, and
remote enterprise connectivity has driven organizations toward adopting Software-Defined Wide Area
Networks (SD-WANs) as a scalable and cost-effective alternative to traditional WAN architectures. SD-
WAN decouples the control plane from the data plane, enabling centralized traffic management, policy
enforcement, and dynamic path selection across multiple transport technologies such as MPLS,
broadband, and LTE. Despite these advantages, SD-WANs continue to face significant challenges in
maintaining optimal routing performance under dynamically changing network conditions.
Traditional routing mechanisms in SD-WANs are typically rule-based or policy-driven, relying on pre-
configured thresholds or static metrics like latency and packet loss. These approaches lack the adaptability
required to respond to real-time variations in network topology, link performance, and application-specific
Quality of Service (QoS) demands. As enterprise networks scale and diversify, routing decisions must

AIJFR25061817 Volume 6, Issue 6 (November-December 2025) 1

 Advanced International Journal for Research (AIJFR)
E-ISSN: 3048-7641 ● Website: [Link] ● Email: editor@[Link]

consider multiple objectives—minimizing latency, maximizing throughput, ensuring reliability, and

maintaining QoS consistency—all while operating within highly dynamic environments. This complexity
motivates the need for intelligent, self-learning routing frameworks that can autonomously adapt to
network fluctuations without manual intervention.

Reinforcement Learning (RL), a branch of machine learning, offers a promising solution to this chal-
lenge. By allowing an agent to learn optimal decision-making policies through continuous interaction with
its environment, RL can dynamically adapt routing strategies based on observed performance outcomes.
When applied to SD-WAN routing, RL enables real-time learning, adaptation, and optimization, lead-
ing to improved network resilience and efficiency. Prior studies have demonstrated the potential of RL in
traffic engineering and congestion control; however, its application in scalable, distributed SD-WAN
environments remains an active area of research.

In this paper, we propose a Reinforcement Learning-Based Adaptive Routing Framework (RLARF)

designed to optimize routing paths in real-time within SD-WAN infrastructures. The proposed framework
models the routing problem as a Markov Decision Process (MDP), where the RL agent continuously
evaluates network states—such as latency, bandwidth, and packet loss—to determine the most efficient
forwarding paths. The system is capable of scaling across large, heterogeneous networks, efficiently
adapting to diverse transport links, and ensuring robust performance even under volatile network condi-
tions.

2. Problem Definition

The increasing adoption of Software-Defined Wide Area Networks (SD-WANs) has revolutionized en-
terprise networking by providing centralized control, dynamic path selection, and efficient utilization of
multiple transport technologies such as MPLS, broadband, and LTE. However, despite these advance-
ments, achieving real-time, intelligent, and scalable routing optimization remains a critical challenge
in SD-WAN environments.

Traditional SD-WAN routing mechanisms typically depend on static policies or threshold-based rules,
where routing decisions are predefined according to fixed metrics like latency, jitter, or packet loss. While
such methods are effective under stable network conditions, they fail to adapt efficiently to highly dy-
namic and heterogeneous network environments, where link quality, bandwidth availability, and ap-
plication-specific Quality of Service (QoS) requirements frequently fluctuate. This lack of adaptability
often results in suboptimal path selection, leading to increased latency, reduced throughput, packet loss,
and degraded application performance.

Moreover, as enterprise networks continue to expand, the scalability and complexity of managing routing
across diverse, distributed infrastructures become significant obstacles. Static routing approaches cannot
efficiently accommodate the continuous learning and decision-making required to optimize performance
across a wide variety of transport links and traffic patterns. Additionally, network unpredictability caused
by congestion, link failures, or changing QoS requirements further exacerbates the problem, emphasizing
the need for an autonomous, data-driven routing mechanism capable of adapting in real time.

AIJFR25061817 Volume 6, Issue 6 (November-December 2025) 2

 Advanced International Journal for Research (AIJFR)
E-ISSN: 3048-7641 ● Website: [Link] ● Email: editor@[Link]

Therefore, the core problem addressed in this research is the lack of an adaptive, intelligent, and
scalable routing framework that can dynamically optimize routing paths in SD-WANs based on
continuous feedback from network states. The difficulty lies in creating a system that can:

1. Continuously learn and adapt to varying network conditions and traffic demands.
2. Optimize routing decisions to minimize latency, maximize throughput, and enhance reliability.
3. Ensure scalability and robustness in large, distributed, and heterogeneous SD-WAN environ-
ments.
4. Incorporate QoS-awareness to meet the diverse performance requirements of real-time and crit-
ical applications.

To address these challenges, this work proposes a Reinforcement Learning-Based Adaptive Routing
Framework (RLARF) that formulates SD-WAN routing as a Markov Decision Process (MDP). By
leveraging reinforcement learning techniques, the framework enables the SD-WAN controller to autono-
mously learn optimal routing policies through interaction with the network environment, thus achieving
self-optimization, resilience, and improved overall network performance.

3. Existing System

Traditional Software-Defined Wide Area Network (SD-WAN) routing systems rely primarily on
static, policy-driven, or rule-based mechanisms to manage traffic across multiple network links. In
these systems, routing decisions are guided by pre-defined configurations set by network administrators,
based on conventional performance metrics such as latency, jitter, packet loss, and bandwidth utiliza-
tion. Although such methods provide a certain degree of control and predictability, they lack the intelli-
gence and adaptability required to respond to rapidly changing network dynamics.

In a typical existing SD-WAN setup, the controller uses threshold-based policies or shortest-path al-
gorithms to select optimal routes. For instance, if latency exceeds a specified limit or packet loss in-
creases beyond a threshold, traffic is rerouted to an alternative link. However, these mechanisms operate
reactively rather than proactively, resulting in delayed responses to network fluctuations. Moreover,
they are unable to anticipate performance degradation or learn from past network behavior, which limits
their efficiency in highly variable environments.

Several enhancements have been made to improve SD-WAN routing efficiency, including heuristic al-
gorithms, multi-path optimization techniques, and application-aware routing policies. These tech-
niques attempt to balance network load and ensure QoS compliance for critical applications. Nonethe-
less, such methods still depend heavily on manual configuration, static rule sets, and periodic moni-
toring, which make them inadequate for large-scale and heterogeneous networks.

Furthermore, existing systems often suffer from the following limitations:

1. Lack of Real-Time Adaptability:

Routing decisions are based on fixed thresholds and do not adapt dynamically to continuous net-
work state changes.

AIJFR25061817 Volume 6, Issue 6 (November-December 2025) 3

 Advanced International Journal for Research (AIJFR)
E-ISSN: 3048-7641 ● Website: [Link] ● Email: editor@[Link]

2. Limited Learning Capability:

Existing routing mechanisms lack the ability to learn from historical data or predict future net-
work conditions, leading to repetitive suboptimal decisions.
3. Scalability Constraints:
Static or semi-dynamic approaches struggle to maintain efficiency as the network grows in size
and complexity, especially across geographically distributed SD-WAN deployments.
4. Inadequate QoS Awareness:
Conventional routing does not effectively differentiate between applications with varying QoS
requirements, causing degradation in performance for latency-sensitive applications.
5. Reactive Behavior:
Most routing protocols respond to changes after degradation has already occurred, rather than
predicting and preventing such issues proactively.

These shortcomings underscore the necessity for a more intelligent, self-learning, and adaptive rout-
ing approach that can make real-time decisions based on continuously evolving network conditions. To
overcome these challenges, this research introduces a Reinforcement Learning-Based Adaptive Rout-
ing Framework (RLARF) that employs machine learning techniques to achieve autonomous optimi-
zation, scalability, and robust routing performance in dynamic SD-WAN environments.

4. Proposed System

To overcome the limitations of traditional rule-based and static routing mechanisms in Software-Defined
Wide Area Networks (SD-WANs), this research proposes a Reinforcement Learning-Based Adaptive
Routing Framework (RLARF). The proposed system aims to enable intelligent, autonomous, and real-
time routing optimization by leveraging reinforcement learning (RL) techniques to dynamically adapt
to changing network conditions and application-specific Quality of Service (QoS) requirements.

4.1 System Overview

The RLARF framework integrates a Reinforcement Learning Agent within the SD-WAN controller
that continuously interacts with the network environment to learn optimal routing strategies. The agent
observes the current network state, such as latency, bandwidth utilization, packet loss, and jitter, and se-
lects an appropriate routing action to maximize overall network performance. Based on feedback re-
ceived from the environment (reward signals), the agent iteratively refines its policy to improve future
decisions. The system operates in a closed-loop learning process, allowing continuous adaptation to
real-time variations in link quality and traffic demands. By modeling routing decisions as a Markov De-
cision Process (MDP), the framework ensures that the RL agent can predict and respond proactively to
dynamic network states, achieving intelligent and scalable routing across distributed SD-WAN infra-
structures.

AIJFR25061817 Volume 6, Issue 6 (November-December 2025) 4

 Advanced International Journal for Research (AIJFR)
E-ISSN: 3048-7641 ● Website: [Link] ● Email: editor@[Link]

4.2 Architectural Components

The proposed system consists of the following major components:

1. SD-WAN Controller:
Acts as the central control plane that manages the network topology, collects link metrics, and
executes routing decisions based on the RL agent’s output.
2. Reinforcement Learning Agent:
The core component of RLARF that learns the optimal routing policy through trial-and-error in-
teractions. It employs algorithms such as Deep Q-Learning (DQN) or Proximal Policy Optimi-
zation (PPO) to map network states to optimal routing actions.
3. Network State Monitor:
Continuously collects real-time metrics, including latency, jitter, bandwidth usage, link loss
rates, and traffic load, from all connected SD-WAN edges and tunnels.
4. Reward Function Module:
Evaluates the performance of each routing decision and provides a reward signal to the RL
agent. The reward is computed based on key QoS metrics such as latency reduction, throughput
improvement, and packet delivery success rate.
5. Data Plane:
Consists of multiple SD-WAN edge devices and transport links (e.g., MPLS, LTE, Broadband).
These nodes implement the routing actions determined by the controller.
6. Policy Repository:
Stores the learned policies and model weights of the RL agent, enabling the system to recall pre-
viously successful routing strategies and improve decision efficiency over time.

4.3 Working Mechanism

The proposed RLARF framework operates in the following sequence:

1. State Observation:
The network monitor gathers real-time parameters such as link utilization, latency, and packet
loss to form the state (S) representation.
2. Action Selection:
The RL agent selects a routing action (A)—choosing the optimal path or adjusting traffic alloca-
tion among available links—based on the current policy.
3. Reward Evaluation:
After executing the routing decision, the system observes the resulting network performance and
assigns a reward (R) proportional to QoS improvement (e.g., reduced latency or increased
throughput).
4. Policy Update:
The agent updates its routing policy using reinforcement learning algorithms to maximize the cu-
mulative reward over time.

AIJFR25061817 Volume 6, Issue 6 (November-December 2025) 5

 Advanced International Journal for Research (AIJFR)
E-ISSN: 3048-7641 ● Website: [Link] ● Email: editor@[Link]

5. Continuous Adaptation:
The framework continuously repeats this process, enabling the network to autonomously adapt to
real-time changes in link conditions and traffic demands.

4.4 Key Features and Advantages

 Dynamic Path Optimization: Learns optimal routes in real-time based on live network data.
 QoS-Aware Decision Making: Ensures application-specific requirements are met for latency-
sensitive traffic.
 Scalability: Efficiently manages routing decisions in large, geographically distributed networks.
 Autonomous Learning: Reduces manual configuration by continuously refining routing strate-
gies.
 Resilience and Reliability: Quickly adapts to link failures and variable network conditions to
maintain consistent performance.

4.5 Expected Outcome

The proposed Reinforcement Learning-Based Adaptive Routing Framework is expected to achieve:

 Up to 30–50% reduction in average latency compared to static routing algorithms.

 Significant improvements in throughput and bandwidth utilization efficiency.
 Enhanced robustness against network volatility and congestion.
 Autonomous self-optimization of routing policies with minimal human intervention.

5. Literature Review

This section reviews prior work relevant to intelligent routing and traffic engineering in SD-WAN/SDN
environments, with emphasis on reinforcement-learning (RL) approaches, QoS-aware routing, scalabil-
ity, and robustness in dynamic networks.

SD-WAN challenges and state of practice.

SD-WAN provides centralized control, multi-transport support (MPLS, broadband, LTE), and applica-
tion-aware policies, which make it attractive for modern enterprises. However, deploying SD-WAN at
scale introduces challenges in real-time routing, QoS enforcement, and resilience under highly dynamic
link conditions—issues that recent surveys identify as open research problems.

Early and classical approaches to SD-WAN routing.

Conventional SD-WAN controllers typically rely on rule/threshold-based policies, shortest-path heuris-
tics, or multi-path load sharing. Extensions based on heuristics, overlay selection, and application-aware
rate allocation improve performance but remain largely reactive and require manual tuning, limiting
their effectiveness in rapidly changing environments. Works proposing SDN-based overlays and control-
ler-centric TE illustrate practical gains while also highlighting scalability and adaptability limits.

AIJFR25061817 Volume 6, Issue 6 (November-December 2025) 6

 Advanced International Journal for Research (AIJFR)
E-ISSN: 3048-7641 ● Website: [Link] ● Email: editor@[Link]

Reinforcement learning for routing and traffic engineering.

Deep reinforcement learning (DRL) has emerged as a promising model-free paradigm for online routing
and traffic engineering. Several studies demonstrate that DRL agents (e.g., DQN, DDPG, A3C/PPO) can
learn routing policies that outperform static baselines (shortest-path, ECMP) on metrics such as latency,
packet loss, and throughput in simulated SDN environments. Notable contributions include ENERO and
related DRL traffic-engineering investigations that show DRL’s ability to optimize complex multi-met-
ric objectives. These works establish DRL as a viable approach for dynamic routing control.

QoS-aware and application-specific RL methods.

Several recent papers explicitly target QoS awareness in RL-based routing. Centralized RL schemes and
hybrid controllers have been proposed to embed QoS objectives (latency, jitter, packet delivery ratio)
into reward designs, sometimes augmented by graph embeddings or neural encoders to capture network
topology. Such approaches demonstrate improvements in application-level performance but often evalu-
ate on small to moderate topologies and under constrained traffic models.

Scalability, multi-agent designs, and heterogeneous links.

To handle larger and distributed networks, researchers have explored multi-agent RL, hierarchical
agents, and decentralized control where edge nodes or local controllers coordinate with a central policy.
Multi-agent setups better map to multi-domain or geographically distributed SD-WAN deployments and
can reduce decision latency, but they introduce coordination and non-stationarity challenges. Mean-
while, studies on adaptive overlay selection and hybrid controller architectures underscore the practical
need to support heterogeneous transport technologies at scale.

Existing SD-WAN-focused prototypes and evaluations.

Several applied studies and technical reports investigate RL agents within emulated SD-WAN testbeds,
measuring gains from RL-based routing under link failures and variable conditions. Industry-oriented
analyses and academic prototypes show promising latency and throughput gains but also point out limi-
tations in training stability, exploration safety (for live networks), and policy transfer across topologies.

Gaps and open problems.

While prior work demonstrates the potential of RL for routing and TE in SDN/SD-WAN, important gaps
remain: (1) most RL studies evaluate on synthetic or small-scale topologies rather than large, heteroge-
neous SD-WANs; (2) few proposals jointly optimize multiple conflicting objectives (latency, through-
put, reliability) in a QoS-aware manner for diverse application classes; (3) robustness under extreme dy-
namics (rapid link flapping, mixed transport characteristics) and safe online exploration in production
networks are under-addressed; and (4) there is limited consensus on scalable architectures (centralized
vs. hierarchical vs. fully decentralized) that balance training complexity, inference latency, and policy
transferability. These gaps motivate the present work, which targets a scalable, QoS-aware RL frame-
work designed specifically for real-time routing in heterogeneous SD-WAN deployments.

AIJFR25061817 Volume 6, Issue 6 (November-December 2025) 7

 Advanced International Journal for Research (AIJFR)
E-ISSN: 3048-7641 ● Website: [Link] ● Email: editor@[Link]

Proposed State Diagram

6. Methodology

The proposed system introduces a Reinforcement Learning-Based Adaptive Routing Framework

(RLARF) designed to dynamically optimize routing paths in real-time SD-WAN environments. The
methodology focuses on integrating intelligent learning agents within the SD-WAN controller to con-
tinuously analyze network conditions, predict link performance, and make routing decisions that mini-
mize latency, maximize throughput, and ensure reliability.

The following subsections describe the step-by-step methodology used to implement and evaluate the
proposed framework.

 System Architecture Overview

 Reinforcement Learning Framework
 Learning Algorithm
 Real-Time Adaptation and Feedback Loop
 Scalability and Multi-Agent Collaboration
 Evaluation Metrics
 Implementation Tools

AIJFR25061817 Volume 6, Issue 6 (November-December 2025) 8

 Advanced International Journal for Research (AIJFR)
E-ISSN: 3048-7641 ● Website: [Link] ● Email: editor@[Link]

7. Summarizing Table

Method/Technique Purpose Equations/Concepts Used

Software-Defined Wide Provides centralized control and dynamic SDN Control Plane, Open-
Area Network (SD- management of WAN traffic across multi- Flow Protocol, Policy-Based
WAN) ple links. Routing
Enables the system to learn optimal routing
Reinforcement Learning Markov Decision Process
strategies through trial-and-error interac-
(RL) (MDP): (S, A, R, P)
tions with the network environment.
Approximates the Q-value function using a Q-Value Update: Q(s,a) ←
Deep Q-Learning (DQL) neural network to make intelligent routing Q(s,a) + α [r + γ maxQ(s',a')
decisions. - Q(s,a)]
Probability-based Action Se-
Balances exploration of new paths and ex-
ε-Greedy Exploration lection: ε (explore) / (1−ε)
ploitation of known optimal routes.
(exploit)
R = w₁(1/Latency) +
Evaluates route efficiency based on latency,
Reward Function w₂(Throughput) − w₃(Pack-
throughput, and packet loss performance.
etLoss)
Enhances learning stability by reusing past Replay Buffer (B) for random
Experience Replay
experiences for training the DQL model. mini-batch updates
Network State Feedback Loop
Dynamically updates routing tables based
Policy Update Module & Continuous Policy Optimi-
on the trained RL agent’s decisions.
zation
Latency (ms), Jitter (ms),
Measures and evaluates routing perfor-
QoS Metrics Analysis Packet Loss (%), Throughput
mance under varying conditions.
(Mbps)

8. Acknowledgement

To everyone who helped make this research a success, we would like to sincerely thank you. First and
foremost, we thank our Head of the Department, Dr. Manju Vyas, for her guidance and invaluable support
throughout the course of this research paper. We also extend our thanks to our colleagues and collaborators
for their helpful discussions and constructive feedback during the development of the models.

9. Authors' Biography

Rahul Guha is a dedicated researcher and educator with a strong academic foundation in computer
science and software engineering. He earned his Bachelor of Technology ([Link]) in Computer Science
from Ajmer Institute of Technology, Ajmer and subsequently completed his Master of Technology
([Link]) in Software Engineering from Bhagwant University. His academic interests include software-

AIJFR25061817 Volume 6, Issue 6 (November-December 2025) 9

 Advanced International Journal for Research (AIJFR)
E-ISSN: 3048-7641 ● Website: [Link] ● Email: editor@[Link]

defined networking (SDN), machine learning for network optimization, and intelligent systems for
enterprise networks. He has published and guided research in these areas, contributing to both curricula
and applied projects in academic institutions.

Nidhi Singh is an academic and researcher with expertise in Machine Learning and Artificial
Intelligence. She holds a Bachelor of Technology ([Link]) in Computer Science from MIT, Kota, and
a Master of Technology ([Link]) from Jagannath University, Jaipur. She is currently pursuing her
Ph.D. from Rajasthan Technical University (RTU), Kota, focusing on the application of machine
learning techniques for intelligent network systems.

Ms. Ambika Bagri is an academician with expertise in Computer Networks and Network Security.
She obtained her Bachelor of Technology ([Link]) from Marudhar Engineering College, Bikaner,
and her Master of Technology ([Link]) from Maharshi Arvind College of Engineering and Research
Center, Jaipur.

Mr. Pankaj Kumar Sharma is an academic professional specializing in Computer Networks and
Communication Systems. He completed his Bachelor of Engineering (B.E.) from Gyan Vihar College,
affiliated with Rajasthan University, and earned his Master of Technology ([Link]) from Jagannath
University, Jaipur.

References
References within Main Content of the Research Paper

1. Troia, S., Sapienza, F., Varé, L., & Maier, G. (2020). On deep reinforcement learning for traffic
engineering in SD-WAN. IEEE Journal on Selected Areas in Communications, 39(7), 2198-2212.
2. Bellini, M. (2022). A Framework for Routing Optimization for SD-WAN using Reinforcement
Learning.
3. Nazemi Absardi, Z., & Javidan, R. (2024). A predictive SD‐WAN traffic management method for
IoT networks in multi‐datacenters using deep RNN. IET Communications, 18(18), 1151-1165.
4. Nazemi Absardi, Z., & Javidan, R. (2024). A predictive SD‐WAN traffic management method for
IoT networks in multi‐datacenters using deep RNN. IET Communications, 18(18), 1151-1165.
5. Iddalagi, P., & Mishra, A. (2024, May). Reinforcement Learning-based Deep Queue Network
Model for Tunnel Health Probing Protocol Management in SD-WAN. In 2024 IEEE International
Conference on Machine Learning for Communication and Networking (ICMLCN) (pp. 467-473).
IEEE.
6. Kumari, L., Guha, R., & Gera, G. Synthesis of Speech & Speaker Recognition Using Data Driven
Approach.
7. Iddalagi, P., & Mishra, A. (2024, May). Reinforcement Learning-based Deep Queue Network
Model for Tunnel Health Probing Protocol Management in SD-WAN. In 2024 IEEE International
Conference on Machine Learning for Communication and Networking (ICMLCN) (pp. 467-473).
IEEE.
8. Abdulghani, A. M., Abdullah, A., Rahiman, A. R., Hamid, N. A. W. A., Akram, B. O., & Raissouli,
H. (2025). Navigating the Complexities of Controller Placement in SD-WANs: A Multi-Objective

AIJFR25061817 Volume 6, Issue 6 (November-December 2025) 10

 Advanced International Journal for Research (AIJFR)
E-ISSN: 3048-7641 ● Website: [Link] ● Email: editor@[Link]

Perspective on Current Trends and Future Challenges. Computer Systems Science &
Engineering, 49.
9. Petrović, T., Vidaković, A., Doknić, I., Veinović, M., & Bojović, Ž. (2025). An Adaptive
Application-Aware Dynamic Load Balancing Framework for Open-Source SD-
WAN. Sensors, 25(17), 5516.
10. Chen, Y., Gu, A., Cui, L., & Lin, L. (2025). MTEAL: Network routing optimization of SD-WAN
traffic engineering integrating multi-dimensional QoS metrics. Journal of Network and Computer
Applications, 104272.
11. Abdulghani, A. M., Abdullah, A., Rahiman, A. R., Abdul Hamid, N. A. W., & Akram, B. O.
(2025). Network-Aware Gaussian Mixture Models for Multi-Objective SD-WAN Controller
Placement. Electronics, 14(15), 3044.
12. Adily, A., & Blessing, A. I. (2024). The Future of Connectivity: Trends in MPLS and SD-WAN:
Exploring emerging trends and technologies that will shape the future of MPLS and SD-WAN
solutions.

AIJFR25061817 Volume 6, Issue 6 (November-December 2025) 11