Application on MobilityFirst Karthikeyan Ganesan, Wuyang Zhang, - - PowerPoint PPT Presentation

Real-Time Cyber Physical Systems Application on MobilityFirst

Karthikeyan Ganesan, Wuyang Zhang, Zihong Zheng Shantanu Ghosh, Avi Cooper

TEAM M MEMBER BERS

WINLAB SUMMER 2015

Karthi hikey eyan an Ganes esan an Wuyang ang Zhang ng Zihong

• ng Zheng

ng Shant antanu anu Ghosh Avi Cooper er

BIG G PICTU CTURE RE OF OUR UR PRO ROJECT JECT

WINLAB SUMMER 2015

Mobil ilityFi ityFirst rst Virtua tual Ne Networ work

Client ent side: de:

Run an insta tance nce of camera ra syste stem; m; Transmi nsmits ts video

• in

standard rd forma mat; t; Simple le graphical hical interf rface ce to display ay results lts

Serv rver r side: e:

Imple lement ment server ver applicatio cation for object ct recogn gniti ition

• n;

Return rn the result lt

CPS Application based on MF

CU CURR RRENT NT FRA RAME

WINLAB SUMMER 2015

Application: Finished the combination of Glass-to-Phone and Phone-to-Node

• communication. The Google Glass is now able sending image to server and

getting back the result. Image Recognition: Implemented the strategy to increase the accuracy and also cut down the mismatch issues. Cloud Computing: All configurations are done. The image recognition program could be run in a certain number orbit nodes side-by-side with STORM framework. Some essential efficiency-related experiments are done.

Some Images From The Google Glass

WINLAB SUMMER 2015

Index: 19 Successfully Matched! Index: 4 Successfully Matched!

STORM

WINLAB SUMMER 2015

Background: As a server processing a large amount of real time requests, requested images will be set into a queue waiting for be processed.. They have to keep waiting before the server completes their previous requests, which delays unnecessary requested time. Requested Images Server

STORM

WINLAB SUMMER 2015 http://blog.gigaspaces.com/wp-content/uploads/image/topology.png

Spout: Water Source (tweets, images, news….) Bolt: Water Processing Machine ( popular tweets ranking, image matching, news follower counting)

STORM Application

WINLAB SUMMER 2015

365,246 182,175 89,825 47,036 100000 200000 300000 400000 Processing Time of 1000 Images Serial 2 Bolts 4 Bolts 8 Bolts Time/ms

Result Analysis

WINLAB SUMMER 2015

Curve tendency ? The relation curve between the number of bolts and processing time is roughly a linear method. Request interval and unit service time? When the request interval is larger than the unit service time, the actual total service time equals to that of the unit service time but without any waiting time in a queue. Unit service time Unit service time Request Interval 365 47 100 200 300 400 Serial 8 Bolts To achieve the service time will not involve the queue waiting time, the serial method allows the minimal request interval is 365ms, but only 47 ms for 8 bolts!! Average Unit Service Time

THE WHOLE CONNECTION

WINLAB SUMMER 2015

Master node as Server manages/allocate image recognition jobs Android phone

MF

WIFI Access Point on Master Node

Generate STORM

Slaves nodes

ORBIT nodes Bluetooth

Google Glass

Combination of CPS server and STORM cluster

WINLAB SUMMER 2015

CPS Server STORM cluster More components…

Spout Bolt Bolt Bolt

Few more bolts… Client

• 1. Send an image &

request result

• 2. Notify the Spout about the new

request

• 7. Return the

result

• 3. Allocate a Bolt to

take this request

• 4. Response the server

and get the image

• 6. Send back the result
• 5. Image Recognition

Ne Next t Week k Plan

WINLAB SUMMER 2015

Do the combi mbination ion of our CPS S server r and the Image ge Recogn gnition ition progra

• gram

m based on STOR ORM. Try to improv rove the stability ility of both the Google gle Class s and Android

• id

pho hone ne progr grams. ms. And of course rse the Post ster! r!

Questions?