Boosting Artificial Intelligence in Software Defined Systems with - - PowerPoint PPT Presentation

boosting artificial intelligence in software defined
SMART_READER_LITE
LIVE PREVIEW

Boosting Artificial Intelligence in Software Defined Systems with - - PowerPoint PPT Presentation

2 ' f National University of Defense Technology Boosting Artificial Intelligence in Software Defined Systems with Open Infrastructure Li Zhou , Qi Tang, Jun Xiong, Haitao Zhao, Shengchun Huang, Jibo Wei College of Electronic


slide-1
SLIDE 1

ý 2 Ñ Ñ ' f

National University of Defense Technology

Boosting Artificial Intelligence in Software Defined Systems with Open Infrastructure

College of Electronic Science, National University of Defense Technology, China Hunan Engineering Research Center of Software Radio, China

Li Zhou, Qi Tang, Jun Xiong, Haitao Zhao, Shengchun Huang, Jibo Wei

Nov 15th, 2018

slide-2
SLIDE 2

ý 2 Ñ Ñ ' f

National University of Defense Technology

Software Defined System

2

software defined radio software defined radar

network cloud

  • 'software defined' is being labeled in a growing

number of systems. terminal

software defined mobile network software defined sensor network software defined satellite network software defined data center software defined cloud computing software defined storage software defined visualization

slide-3
SLIDE 3

ý 2 Ñ Ñ ' f

National University of Defense Technology

Software Defined System

3

  • Software infrastructure plays an essential role.

Type Commercial Open Source Open Standard Terminal

Windows, iOS Linux, Android, ROS SCA, STRS

Network

Cisco ACI, VMware NSX Beacon, Floodlight, OpenDaylight, ONOS OpenFLow

Cloud

Microsoft Azure, Amazon Web Services (AWS), Alibaba Cloud OpenStack, Kubernetes

Hardware Software Infrastructure

Unified or standardized APIs Components, Applications, Services

slide-4
SLIDE 4

ý 2 Ñ Ñ ' f

National University of Defense Technology

Software Defined System

4

Computing Resource Pool Storage Resource Pool Network Resource Pool

Kubernetes Docker Ceph User Management Servers Switches Routers SDRs

Hardware Layer Infrastructure Layer Service Layer

Software Repository Application Develepment SDR Platform Controller SDR Platform Monitor Test and Certification Registry Identity ...

CK CK CK P P CK P CK P

Terminal SDS Cloud SDS

Infra API

General Processors CPU GPU FPGA ... Hardware Devices RF USB Ethernet ...

...

Application 1

Comp Comp Comp Comp Comp

Application N

Comp Comp Comp Comp Comp

Application 2

Comp Comp Comp Comp Comp

Infrastructure Layer Operating System

Comp Comp Comp Comp

Application Layer Hardware Layer

e.g. SCA

  • Deployment, management, interconnection

and intercommunication of software components

e.g. Kubernetes

  • Automate deployment, scaling, and

management of containerized services

slide-5
SLIDE 5

ý 2 Ñ Ñ ' f

National University of Defense Technology

AI-able Terminal SDS

5

HAL API Infra API Act API

General Processors CPU GPU FPGA ... Hardware Devices RF USB Ethernet ...

Cognitive Engine

Observe API

...

Application 1

Comp Comp Comp Comp Comp

Application N

Comp Comp Comp Comp Comp

Application 2

Comp Comp Comp Comp Comp

Infrastructure Layer Operating System

Comp Comp Comp Comp

Application Layer Hardware Layer Orient Decide Training Knowledge ...

AI-able Terminal SDS

  • Cognitive Engine: OODA Loop & Training
  • Possible Hardware: CPU, GPU, AI Chips
  • Observe:

– query data from applications – spectrum data, parameters, etc.

  • Act:

– configure parameters into applications – Install/uninstall/start/stop/switch applications

Cognitive Engine Application Repository Sensing Application Waveform Application Transceiver Wireless Environment

  • bserve

act act

  • bserve

How to boost AI in SDSs with open infrastructure?

  • Inspired by NASA Glenn cognitive communications systems project
slide-6
SLIDE 6

ý 2 Ñ Ñ ' f

National University of Defense Technology

SDSDevOps Environment

Cloud Platform

Developments/Operations Terminal SDSs Intranet/ Internet

slide-7
SLIDE 7

ý 2 Ñ Ñ ' f

National University of Defense Technology

Open AI-able Cloud Infrastructure

7

Computing Resource Pool Storage Resource Pool Network Resource Pool

Kubernetes Docker Ceph User Management Servers Switches Routers SDRs

Hardware Layer Infrastructure Layer Service Layer

Software Repository Application Develepment SDR Platform Controller SDR Platform Monitor Test and Certification Learning Framework Data Analysis ... Registry Identity ...

CK CK CK P P CK P CK P

Open AI-able Cloud Infrastructure

  • User Management
  • Software Repository
  • Application Development
  • SDR Platform Controller
  • SDR Platform Monitor
  • Test and Certification

Basic Services

Cloud

  • Learning Framework:

‒ TensorFlow, ‒ Caffe, ‒ PyTorch ‒ …

  • Data Analysis:

‒ Pandas, ‒ Statsmodels ‒ scikit-learn ‒ … AI as a Service (AIaaS)

slide-8
SLIDE 8

ý 2 Ñ Ñ ' f

National University of Defense Technology

Open AI-able Cloud Infrastructure

8

  • Create training scenarios

‒ Centralized networks ‒ Decentralized networks ‒ Distributed networks ‒ Wired/wireless backhaul

  • Build dataset

‒ Geography data ‒ Spectrum data ‒ Channel data

  • Train parameters

‒ Neural network ‒ …

  • Train policies

‒ Support Vector Machine ‒ Reinforcement Learning ‒ …

  • Knowledge database

‒ Transfer learning

Optimize cognitive engines in SDSs

slide-9
SLIDE 9

ý 2 Ñ Ñ ' f

National University of Defense Technology

Deployment on NUDT Campus

9

Indoor SDRs Outdoor SDRs

  • 60 SDRs are deployed initially

‒ 24 SDRs in the laboratory ‒ 36 SDRs outside in the campus

  • Cloud servers are deployed in the laboratory

‒ Intranet deployment only currently ‒ Internet deployment is on the way Cloud Servers

slide-10
SLIDE 10

ý 2 Ñ Ñ ' f

National University of Defense Technology

10

Example Application (1)

  • L. Zhou, Z. Sheng, L. Wei, X. Hu, H. Zhao, J. Wei & V. C. Leung, “Green Cell Planning and Deployment for Small Cell Networks in Smart

Cities”. Ad Hoc Networks, vol. 43, pp. 30-42, June 2016.

e.g. Cell Deployment & Planning

  • Assumptions

‒ Spatial traffic (user distribution) varies from time to time in a region following some traffic patterns. ‒ Centralized controller & wired backhaul

  • Problems

‒ How to deploy minimal number of small cell base stations to meet users’ spatial traffic requirements? ‒ How to control the ON/OFF status

  • f the small cell base stations to

maximize energy efficiency?

slide-11
SLIDE 11

ý 2 Ñ Ñ ' f

National University of Defense Technology

11

Example Application (1) e.g. Cell Deployment & Planning

Traffic Clustering Cell Deployment & Planning

  • Our approach: Support Vector Machine (SVM) + Deep Neural Network
slide-12
SLIDE 12

ý 2 Ñ Ñ ' f

National University of Defense Technology

12

R

CH1 CH2 CH3

  • Problem: How to group users into clusters with optimal size?

Example Application (2)

  • Our approach: Reinforcement Learning (POMDP) + Deep Neural Network

e.g. User Clustering

  • Assumptions

‒ Mobile ad hoc network ‒ Users are divided into clusters distinguished by spectrum band (i.e., channels). ‒ Too large cluster size would result in heavy intra-cluster communication collisions ‒ Too small cluster size would result in large number of clusters, which leads to complex inter-cluster communication

slide-13
SLIDE 13

ý 2 Ñ Ñ ' f

National University of Defense Technology

13

Summary and Future Work

  • SDSDevOps is an environment that devotes to the

development, operation, test and training of SDSs for students and researchers based on an open cloud infrastructure.

  • Massive centralized/Decentralized/distributed, wired/wireless

backhaul scenarios can be created.

  • We plan to deploy and validate more applications that we used

to study them by simulation on SDSDevOps. ‒ Cooperative Spectrum Sensing ‒ Cognitive MAC protocols ‒ Resource Management ‒ Mobility Management ‒ …