AI Agent in Software-Defined Network: Agent-Based Network Service Prediction and Wireless Resource Scheduling Optimization

AI Agent in Software-Defined Network: Agent-Based Network Service Prediction and Wireless Resource Scheduling Optimization

Table of Contents


With the development of a Software-defined Network (SDN), there will be a large number of devices to access the network, which may cause an incalculable burden to the communication network. In addition, due to the high-bandwidth in the 5G era, innovation will occur in different fields. There are not only strict requirements on the communication capability ofSDN for these application scenarios but also a lot of computing resources. For massive access devices, it is difficult for the traditional service resource scheduling and allocation system to meet user demand growth. To address the above-stated problems,an artificial intelligence agent (AI Agent) system is put forth in this paper. AI Agents can be deployed in different layers of theSDN, thus realizing functions like network service prediction and resource scheduling. A brand new AI Agent framework is designed, and an AI algorithm is adopted to replace the traditional service prediction and resource scheduling strategies.In the meantime, a relevant agent deployment scheme is put forward. Finally, an AI Agent-based simulation experiment for resource scheduling is designed, and the accuracy in-network service prediction and rationality in resource allocation based on this framework are tested. The experiment result showed that the operation efficiency of the SDN can be effectively improved, and the resource hit ratio and user service quality may be improved with AI-agent-based traffic prediction and resource allocation model.

  • Author Keywords

    • Artificial intelligence agent (AI Agent) ,
    • software-defined network (SDN) ,
    • traffic prediction ,
    • wireless resource allocation
  • IEEE Keywords

    • Artificial intelligence ,
    • Resource management ,
    • 5G mobile communication ,
    • Scheduling ,
    • Prediction algorithms ,
    • Base stations ,
    • Communication networks


The fifth generation of the mobile communication system (5G) will soon become a reality, and this a significant breakthrough based on the existing communication network. According to current research in different countries, the peak5G technology rate will increase by tens of times compared to4G, from 100Mb/s to dozens of Gb/s [1]. Secondly, the end-to-end delay will decrease from more than ten milliseconds in the4G era to several milliseconds [2] [3]. It can also be expected that network access of a large number of devices could cause an incalculable burden to the communication network after5G technology construction is completed. Thirdly, due tothe high-bandwidth in the 5G era, innovation will occur indifferent fields, such as ultra-high-definition video caching [4],ultra-high-resolution game rendering, telemedicine [5], and automatic driving [6]. For these application scenarios, there arenot only strict requirements on the communication capability of the communication network, but also a lot of computing resources [7] [8]. The network access of massive devices,decreasing the in-network communication delay, and the op-timal allocation of resources will be the three objectives for enhancing 5G communication quality [9].In the 5G era, the application of Software-defined Net-work(SDN) will be more extensive and important. The sep-aration of control layer and forwarding data layer in SDNnetwork makes the network more open and supports the upperapplication flexibly. Based on this, packet data connection,traffic identification, variable QoE, downlink buffer, packetconversion and selective link can be realized [10] [11].User mobility, randomness, and service diversity will re-sult in an unbalanced load on the network [12]. Therefore,mobile service prediction and wireless resource scheduling ina mobile communication network have captured the interestof researchers. To allocate reasonable mobile network serviceresources (such as downloading, response, and storage) todifferent users in different application scenes, and to conductreasonable allocation to network resources on the basis ofsatisfying as many user groups as possible. As for traditionalnetwork prediction, experts analyze a user’s data servicerequests over a period of time and establish the network trafficprediction model with prior knowledge in the communicationnetwork. On the one hand, this mode requires the support of apowerful expert system, and the modeling must be conductedby professionals, which hinders the progress of data analysisin the general communication network; on the other hand,due to the heterogeneity, complexity and dynamicity of thecurrent communication network, there will be node access inthe network or node exits out of the network at any time, andthus data modeling is not sufficient to accurately describe thereal situation of the network, and it may also cause networkmanagers to make wrong decisions [13]. In addition, with theincreasing demands of the internet, and faced with massiveaccessing devices, the traditional service resource schedulingand allocation system can hardly meet the increasing demandsof users [14] [15].

An agent is a computer program that simulates humanbehavior and relationships with a certain intelligence andcan run autonomously to provide corresponding services. Anagent can function autonomously and consistently, and theyare characterized by residency, reactivity, sociality, initiative,and so on. Agents can be adapted to a certain environment toact spontaneously according to the characteristics of theenvironment, thus achieving the design purpose of an agen-t [16] [17]. In the meantime, the information from environmentthrough sensors can be perceived by agents, which also exertcertain influences on the environment. The introduction of anagent module in the communication system can efficientlysolve many problems in the current communication system.For example, with an agent, user service requests may beanalyzed using neural network algorithms for classification;a long short term memory (LSTM) network may be adoptedto explore the correlation of wireless communication resourceallocation; and a support vector machine (SVM) may be usedto extract and predict the features of different kinds of networkservices.At present, in the field of SDN, an agent-based intelligentperception has been proposed to realize real-time communi-cation environment perception and dynamically allocate com-munication resources and coordinate communication servicesat each terminal [18] [19]. Bikov et al. proposed that amulti-agent learning algorithm should be adopted to conductindependent real-time resource allocation optimization in the5G mixed network framework and that resource allocationoptimization should be achieved independently in each basestation [20]. A kind of communication mechanism amongmultiple agents was proposed by Qi et al. For cooperativework among multiple agents when necessary, thus completinginformation sharing [21]. To address traffic prediction, amobile-agent-based distributed variational Bayesian algorithmwas designed by Mohiyeddin et al., and the traffic density pre-diction of the sensor network is realized with the algorithm bythem [22].In a 5G environment, the peak size of the resourcesrequested by users will be increased from M level to G level;therefore, to improve the communication experience of users,cooperative work of agents and optimal resource allocationscheme should be considered. The allocation process of largeresource blocks and the collaboration of multiple agents inresource allocation under this scene are not considered inthe existing research; for example, in a content distributionnetwork, a large number of users request resources of differentsizes, and the optimal response strategies greatly improve theresource hit ratio and user experience.

Based on the above-stated problems, an artificial intelli-gence agent (AI agent) system is proposed in this paper, AIagents that are deployed at different levels of SDN are adoptedto complete network service prediction, resource scheduling,and other functions. AI agents at different levels play differentroles. An AI agent at the user device level is responsiblefor integrating users’ historical network data, and then theuser data usage behavior patterns can be analyzed so as tomake user resource request predictions. An AI agent at thebase station level is used to aggregate traffic and service typerequest data and to conduct resource scheduling and allocationwith the goal of maximizing the quality of experience (QoE)for users. An AI agent at SDN controller is used to collectand analyze service request data in order to promptly deploypackets with the high request rate on the server of the basestation.


In this paper, a system is designed with AI mobile agentin SDN. The limitations and many problems in the resource allocation of the existing communication network are summarized, and a new wireless resource allocation and scheduling algorithm is proposed. The experimental results showed that the algorithm performs better in resource allocation, and thus better accuracy and lower communication delay could be obtained.

Future work will be carried out focused on three aspects. The first aspect is the security problem, which is the most important problem for mobile agents [41]. We will study how to guarantee the security of data carried in terms of user demand. The second aspect is regarding synergy among intelligent agents. In the simulation experiment, it was assumed that the communication between each base station and the user was unimpeded. However, the actual communication environment was complicated. In addition, the combination of mobile agents and AI may bring benefits, but the lack of information communication would cause information loss. The next step is to investigate the interactions of processes in an amulti-agent environment. The third aspect is user mobility. To simplify the problem, it was assumed that the transition fora mobile range of communication users was always smooth between communities. While in actual scenes, the movement of a user is always complicated and random. To generalize our model, an in-depth study on user mobility must be carried out

About KSRA

The Kavian Scientific Research Association (KSRA) is a non-profit research organization to provide research / educational services in December 2013. The members of the community had formed a virtual group on the Viber social network. The core of the Kavian Scientific Association was formed with these members as founders. These individuals, led by Professor Siavosh Kaviani, decided to launch a scientific / research association with an emphasis on education.

KSRA research association, as a non-profit research firm, is committed to providing research services in the field of knowledge. The main beneficiaries of this association are public or private knowledge-based companies, students, researchers, researchers, professors, universities, and industrial and semi-industrial centers around the world.

Our main services Based on Education for all Spectrum people in the world. We want to make an integration between researches and educations. We believe education is the main right of Human beings. So our services should be concentrated on inclusive education.

The KSRA team partners with local under-served communities around the world to improve the access to and quality of knowledge based on education, amplify and augment learning programs where they exist, and create new opportunities for e-learning where traditional education systems are lacking or non-existent.

FULL Paper PDF file:

AIAgent in Software-defined Network:Agent-based Network Service Prediction andWireless Resource Scheduling Optimization



Y. Cao, R. Wang, M. Chen and A. Barnawi,




AI Agent in Software-Defined Network: Agent-Based Network Service Prediction and Wireless Resource Scheduling Optimization,

Publish in

in IEEE Internet of Things Journal, vol. 7, no. 7, pp. 5816-5826, July 2020,



PDF reference and original file: Click here

+ posts

Somayeh Nosrati was born in 1982 in Tehran. She holds a Master's degree in artificial intelligence from Khatam University of Tehran.

Website | + posts

Professor Siavosh Kaviani was born in 1961 in Tehran. He had a professorship. He holds a Ph.D. in Software Engineering from the QL University of Software Development Methodology and an honorary Ph.D. from the University of Chelsea.

Website | + posts

Nasim Gazerani was born in 1983 in Arak. She holds a Master's degree in Software Engineering from UM University of Malaysia.