Raja Sengupta Wireless Networks Forcars

8/8/2019 Raja Sengupta Wireless Networks Forcars

1/98

Wireless Networks for Cars

CITRIS Research Exchange

Raja Sengupta

Associate Professor

CEE: Systems EngineeringEE: Systems Engineering

University of California at Berkeley, USA

1


2/98

October 21 1999

The record in this proceeding overwhelmingly supportsthe allocation of spectrum for DSRC based ITS

,consumption and pollution, and continue to advance the

nations economy., , , -

Amendment with licensing rules in December 2003

75 MHz of spectrum at 5.9 GHz

2

CEE: Systems UCB


3/98

http://www.its.dot.gov/initiatives/ initiative9.htm

Goal: Achieve nationwide deployment of a communicationsinfrastructure on the roadways and in all production vehicles and to

bl b f k f d i l i h ldnable a number of key safety and operational services that wouldtake advantage of this capability.

systems developed under the IVI and the availability of radio spectrumat 5.9GHZ recently approved by the FCC for Dedicated Short RangeCommunications. The VII would enable deployment of advanced

ve c e-ve c e an ve c e- n rastructure commun cat ons t at coukeep vehicles from leaving the road and enhance their safe movementthrough intersections. These deadly roadway scenarios account for

'

3

CEE: Systems UCB


4/98

The basic idea

Get the radios on the road

Move lots of useful information around

4

CEE: Systems UCB


5/98

Useful Information

http://www.vehicle-infrastructure.org/applications/

5

CEE: Systems UCB


6/98

Traffic ManagementGPS data from cars at

6

CEE: Systems UCB


7/98

7

CEE: Systems UCB


8/98

8

CEE: Systems UCB


9/98

SafetApplicationSpace

Dela : 50 - 100msec

Range:

Ref: VSCC Final Report

9

CEE: Systems UCB


10/98

Cooperative Collision Warning

Berkele rotot e 2004

10

CEE: Systems UCB


11/98

Message size = 200 msecRange = 150 metersVehicle density/lane = 30 meters/vehicleLanes = 8Interference range = 2* transmission range

= Omni-antennas

With perfect schedulingve c e can ransm every msec

Slotted Aloha efficiency = 36%

11

CEE: Systems UCB


12/98

12

CEE: Systems UCB


13/98

13

CEE: Systems UCB


14/98

Useful Information

http://www.vehicle-infrastructure.org/applications/

Tough!One to one transactionsLow latency

14

CEE: Systems UCB


15/98

The community tried to standardize solutions

to new networkin roblems

-

channel 1 Spread the usersChoose SSIDAssociate

DELAY!

Spread the applications

15

CEE: Systems UCB channel 2

Schedule

New security model


16/98

Initiall thin s moved fast .

S ectrum for DSRC allocated in 1999:

http://www.fcc.gov/Bureaus/Engineering_Technology/News_Releases/1999/nret9006.html

an ar za on ase on . a n :

Called 802.11a_Roadside then. Now IEEE 802.11p

http://findarticles.com/p/articles/mi_m0EIN/is_2001_August_30/ai_7769672

Atheros released its first 802.11a chipset in 2001:

http://findarticles.com/p/articles/mi_m0EIN/is_2001_Sept_5/ai_77846651

Triband chip set: 2.4, 5, 5.9 GHz

Support for 10 and 20 MHz channels

-

16

CEE: Systems UCB


17/98


18/98

Status of RSE

VII California Testbed VII California Testbed

18

CEE: Systems UCB www.viicalifornia.org


19/98


20/98


21/98

Something happened while USDOT, Car

com anies have been doin DSRC.

Internet

21CEE: Systems UCB


22/98

Oops! Theres an elephant in the

room..

Internet

22CEE: Systems UCB


23/98

MOBILE MILLENNIUM

http://www.youtube.com/watch?v=TOi5nGCJ1YIhttp://lagrange.ce.berkeley.edu/media/media-coverage.htmlhttp://abclocal.go.com/kgo/story?section=news/technology&id=6227534

Alex BayenAssistant Professor, Systems Engineering,Civil and Environmental Engineering, UC Berkeley

Quinn JacobsonResearch Leader, Mobile Internet Services SystemsNokia Research Center Palo Alto


24/98

Similar notoriety with UAVs

http://blog.wired.com/defense/2008/06/video-iphone-.

Video: iPhone Controls Robot Plane SquadB Noah Shachtman June 16, 2008 9:16:00 AMCate ories: Drones, Video FixSince 2004, a team of professors and students from the University of California,Berkeley has searched for ways to let a single human supervise a team of robot

planes. Now, this Center for Collaborative Control of Unmanned Vehicles has a.In a video taken from this month's Teaching & Technology conference, theBerkeley crew uses an iPhone to pick tasks for its drone squadron, input a set ofcoordinates for a local reconnaissance mission, and send the planes new orders

while the aircraft are in the sky.But don't tell Steve Jobs how the Berkeley folks are using his gadget. Accordingto the terms of the Apple Software Developer Kit agreement, "applications may

- autonomous control of vehicles, aircraft, or other mechanical devices; dispatch or

fleet management; or emergency or life-saving purposes."


25/98

The Mobile Internet..

Macworld | Analysts: iPhone driving smartphone sales

Its also Quik and Twitter and Pownce and MTV

The iPhone has had a profound effect on the smartphone market, panelists at the Smartphone Summit said on Monday.www.macworld.com/article/132769/2008/03/iphone_market.html - 47k - Cached - Similar pages -Note this

Sm artphone sales skyrocket | Tech news blog - CNET News

Smartphones are hot and getting hotter--and are outpacing the sale of laptops, according to market

research firm In-Stat. Read this blog post by Marguerite ...news.cnet.com/8301-10784 3-9816072-7.html - 84k - Cached - Similar a es - Note this _

Smartphone Sales To Triple, Good News For Apple, RIM, Carriers

Mar 20, 2008 ... Smartphone sales are projected to more than triple worldwide in the next six years,from 10% of the overall mobile phone market last year to ...

h lwww.a ey ns er.com smartphone _sales _to_tr p e_goo _news_ or_app e_r m_carr ers -- Cached - Similar pages - Note this

Gartner Says Worldwide Smartphone Sales Grew 29 Percent in First ...

25CEE: Systems UCB

, ... . , .from the first quarter of 2007, according to Gartner, ...www.gartner.com/it/page.jsp?id=688116 - 30k - Cached - Similar pages - Note this


26/98

While DSRC has been trying to put

wireless in the car..

+= nthe car!

26CEE: Systems UCB


27/98

27CEE: Systems UCB


28/98

SAFE TRIP Inspiration: SmartWay 2007

second generation ITS in Japan 1,650 Exhibition attendees; 35 media outlets Ongoing Tokyo field test; 60 vehicles

-infrastructure: 163 TMCs; 2,800 cameras; 180,000 traffic

sensors;

(VICS) with 33,000+ communications beacons; FMradio

Builds upon installed base of ITS users: . 28.3 Million navigation units; 19.6 Million VICS

Motivated by distinctive Japanese conditions Complex street addressing and navigation

Serious motor vehicle crash rate 5X U.S Vehicle density 10X U.S.; congestion / limited

parking $5.60 / gallon fuel; ~$0.40 / mile expressway tolls


29/98

The New A roach The Old Wa

V2V Safety Intersection Tolling Probe Data

Internet: IP, Multicast, IPSec, WPA, http

29CEE: Systems UCB


30/98

Damn these Europeans.How do they

et it ri ht?

30CEE: Systems UCB


31/98

The New A roach The Old Wa

V2V Safety Intersection Tolling Probe Data

Internet: IP, Multicast, IPSec, WPA, http

DSRC may rise of fall Information survives

31CEE: Systems UCB

Lots of inefficiencies, no QoS, waste bandwidthBut its out there :(


32/98


33/98

e onnec e rave er rogram

Caltrans, USDOT-RITA, MTC

Two projects

Mobile Millenium: Berkeley, Nokia, Navteq GEMS Networked Traveler: Berkekey, Nissan,

Navteq

33


34/98

Get yourself a smartphone and go to

www.networkedtraveler.or Demo onl

information youwant to put in

And what youwant to take

out.

34CEE: Systems UCB


35/98

Get yourself a smartphone and go to

www.networkedtraveler.or

information youwant to put in

And what youwant to take

out.

35CEE: Systems UCB


36/98

Then go out into the world with yoursmar p one n your poc e ..

36


37/98

Plan our tri

Iphone movie

37

CEE: Systems UCB


38/98

Rich view of commuting conveniencer p p ann ng s mu -mo a mu -cr er a

38

CEE: Systems UCB


39/98


40/98

If ou are ridin the bus

40

CEE: Systems UCB


41/98

Your Smartphone based personal mobility

assistant sta s tri aware

41

CEE: Systems UCB


42/98

Transit Signal Priority .

42

CEE: Systems UCB


43/98

If you are driving

Quasi-dynamic

43

CEE: Systems UCB


44/98

If you chose pedestrian alerts on your networked traveler

.

Watch out for me!

44

CEE: Systems UCB


45/98


46/98

The Networked Traveler Concept:

.

The Pitch?

Welcome to the world of the Networked Traveler! An easy way to putin and take out information for smarter personalized travel. Get asmartphone and browse www.networkedtraveler.org.

46

CEE: Systems UCB


47/98

What will ha en to DSRC VII?Many of its local information dissemination applications will be realizedby the Mobile Internet Traffic density, parking, accident reports, .

So will Soft safety / Situation Awareness Signage, Careful stopped traffic ahead

However Smartphones get WiFi into cars Is the ad-hoc mobile network finally coming to life?

Ride a new convergence of the mobile Internet and car industry

Will ad-hoc information protocols, hotspot protocols, DSRC chipsetsgo into smartphones

C ti C lli i W i t ki i d d

47

CEE: Systems UCB

Cooperative Collision Warning, cars tracking cars, remains a deep andinteresting application not for the mobile Internet


48/98


49/98

The END

Collaborators

Jim Misener, Co-Investigator GEMS Dr. Susan Dickey, Manager Software Group

CEE: Systems Prof. Alex Ba en

Students: Ching-ling Huang, Shahram Rezaei, ChristianManasseh, Jerry Jariyasunant, Dan Work

49

Prof. Kannan Ramanchandran


50/98

50

CEE: Systems UCB

The community has been working on


51/98

The community has been working onthese roblems for four ears

We need some visible successes to attract people, companies,funding (NSF), etc

The registration information and papers in this field stays roughly the same sizeover years

WHY? Some google trend info (please remove or revise )

51

CEE: Systems UCB


52/98

Examples of highly visible activities (successes ?)

autonomous vehicle capable of driving in traffic ( Link, video clips ),

Mobile millennium Use 3G to transfer real time sensing results from cars on highway, and to deliver real

time info to vehicles (video ).

http://traffic.berkeley.edu/

severa me a an news c ps on: p: ra c. er e e .e u newsre eases. m

Japan Smartway demostration htt ://www.smartwa 2007. /en/index.html

52

CEE: Systems UCB

What are the large programs dedicated


53/98

What are the large programs dedicatedto these roblems?

Japan: Smartway

Europe: CVIS consortium, Cooperative Vehicle Infrastructure Systems ( video )

US: VII (link) Expect to install 250,000 APs

. .

Are these good enough?

53

CEE: Systems UCB


54/98

Sales of 802.11a b chi sets

Its rowin ex onentiall since 2003: huge success: Atheros alone surpassed 1,000,000 shipment per month http://findarticles.com/p/articles/mi_m0EIN/is_/ai_105906482

Progress and acceptability of two things created at the same time . 802.11p (DSRC) for VANET: Failure

54

CEE: Systems UCB


55/98

In dan er of bein obsolete?

New technologies are coming! Higher data rate, wider comm. range

802.11n (MIMO), 802.16 (WiMAX), 802.15.3 (UWB)

Many applications are moving directly towards newer technologies,bypassing DSRC

802.11p and ad hoc setting is not attractive? Are we working on a to-be-obsolete technology?

55

CEE: Systems UCB


56/98

Obsolete in A lications

Gathering real-time Traffic information

Parking information

we can o ose eas y us ng , w y o we nee . p

Some applications have been realized using other technologies

What is left for us to do? Safety applications needs to work in ad hoc setting; its not likely to be

replaced by others.

56

CEE: Systems UCB

What can not be replaced by other


57/98

What can not be replaced by othertechnolo ies

Ad hoc networking to support real time information exchange

Scalability No infrastructure is required

Low latency requirement

Safety applications need as much resource as possible Do we know whats the limit given the current channel allocation and

technology?

57

CEE: Systems UCB


58/98

We don t know how bi the i e is?

The capacity of the channel depends on the location, traffic condition,etc.

This is a multi-accessed channel Shared by all the cars/nodes

What we learn from TCP is that we can still do well without knowing

The sender can adapt its rate to utilize the channel as much as possible

All users can share the pipe fairly

58

CEE: Systems UCB

Shannon s capacity bound and Turbo coding


59/98

Shannon s capacity bound and Turbo coding 1 2

59

CEE: Systems UCB

Shannon s capacity bound and Turbo coding


60/98

Shannon s capacity bound and Turbo coding 2 2

100 SNR vs. BER for rate 1/2 codes

10 -1ear Rate Code SNR Required for BER< 10 -5

-3

10-2

B E R Iterative

Code

Uncoded1948 SHANNON 0dB

1967 (255,123) BCH 5.4dB

1977 Convolutional Code 4.5dB

10-4

Conv. CodeML decoding

CapacityBound

" -

1993 Iterative Turbo Code 0.7dB

2001 Iterative LDPC Code 0.0245dB

0 1 2 3 4 5 6

SNR

. . ,Codes," IEEE Trans. Comms., Vol.44, No.10, Oct 1996.

60

CEE: Systems UCB

nce we now w ere e oun s, wecan work on approaching to it!

Sahai s anytime capacity and its relationshipi h ki d i l


61/98

with tracking a dynamical system

Shannons capacity definition is not sufficient for us to judge whethera channel can su ort trackin of a d namical s stem

A new definition from A. Sahai tells us what should be the notion of

capacity when the channel is used in control loop.

61

CEE: Systems UCB

Definition of Anytime Reliability and


62/98

An time Ca acit

Error probability is graduallydecreasing with time

(Message accuracy isincreasing with time)

Anytime Capacity

62

CEE: Systems UCB

Anytime Capacity for BEC


63/98

y p y

Shannons capacity

eP

2log

63

CEE: Systems UCB


64/98

U if i


65/98

Uniform source rate assum tion

Each node has homogeneous source producing the same information rate

Various simulation is done to explore what rate is possible given current

Message dispatcher idea to reduce message amount

but it still assume homo enous traffic from each node

65

CEE: Systems UCB

R i f N d i l


66/98

Review of Non-ada tive results

Throughput increases before hitting the capacity

Then it decreases fast after reaching capacity

If yes, we can all go home!

If no, how good it can be?

66

CEE: Systems UCB Bianchis 802.11 analysis

Slotted ALOHA to understand the availabled h h


67/98

er node throu h ut

Simple example of what information is available to every node The information amount vanishes to zero as node number increases

Per node throughput vanishes to zerosenders 0)1( 1 nenn

n

receiver

67

CEE: Systems UCB

%361

)1

1(1

en

n

Po er control to share the channel


68/98

Power control to share the channel

Reduce the number of nodes sharing the channel

A common approach is to reduce comm. range so that appropriate amount

Information exchange happens locally since neighboring cars are more physicallyinvolved than remote ones

68

CEE: Systems UCB

Power control from Hartenstein


69/98

Power control from Hartenstein

Assum tions : The Carrier Sense Range is monotonic

All nodes have the same beaconing frequency and packet sizes All nodes have the same Maximum Transmission Power Pmax

Fair Power Control: All vehicles in a certain (busy) area must restrict their beacons transmission power

Maximize the minimum transmission power

Maximize individual coverage

Above combines with various channel models

69

CEE: Systems UCB

Rate adaptation (based on estimation error)can be hel ful


70/98

can be hel ful

ources pro uce t me-vary ng rates

Especially for vehicle safety (estimation purpose) Rate of the source depends on the estimation error of the dynamic process

Communication happens only when it is necessary to release information

This enables multiplexing gain and better sharing the precious channelcapacity

70

CEE: Systems UCB

Fixed Rate Communication


71/98

SVSelf Estimator

OV o r

kSelf Estimator

(Kalman Filter) N e t wFixed Periodic (Kalman Filter)

Sources produce the same amount of information

Communicate even when it s not necessary

71

CEE: Systems

UCB

Variable Rate Communication


72/98

SV OV k

Measurements

e st mator(Kalman Filter)

(Kinematic Eq.s)

N e t w o

Scheduler

Remote Estimator(Kinematic Eq.s)

Neighbor Estimators(Kinematic Eq.s)

NE 1 NE 2 NE n. . .

OV: Other Vehicle(s)

Sources communicate when it s necessary to release information

72

CEE: Systems

UCB

Averaged Tracking Error:Variable Rate vs. Periodic


73/98

73

CEE: Systems

UCB

Tracking Error tail probability:Variable Rate vs. Periodic


74/98

74

CEE: Systems

UCB

Averaged Tracking Errors:Variable Rate with Repetition


75/98

75

CEE: Systems

UCB

Tracking Error tail probability: VariableRate with Repetition


76/98

76

CEE: Systems

UCB

Error-dependent policy (1/3):MSE lower bound


77/98

ac no e ec es s x. pro a y: Comm. rate: Error sensitivity: x. pro a ty:

Lower bound of MSE: Achieved by all renewal-type comm. policies

(Round-robin MSE)Necessary condition for optimal sensitivity

-

77

CEE: Systems

UCB

arrival time equals the number of nodes.

Error-dependent policy (2/3):Sensitivit to error


78/98

O timal sensitivit Matlab simulation results

78

CEE: Systems

UCB

Err-Coll-Dep policy (1/2):formulation


79/98

Improved version of Error-dependent

Use channel collision as indicator to infer the channel conditionEach node decides its Tx. probability: Comm. rate: Error sensitivity:

Collision sensitivity: Tx. probability:

,

79

CEE: Systems

UCB

Err-Coll-Dep policy (2/2): comparisonwith Err-De


80/98

Matlab simulation results

80

CEE: Systems

UCB

O timal collision sensitivit : beta


81/98

~urns ou e a s op ma va ue: , independent of node number!

During congested channel:,

When beta >30, too much backoff

Beta=9 or 39 Beta=29

M

S

M

S

81

CEE: Systems

UCB

ra e ra e

Short Summar


82/98

-,is the right technology to work on

For active safet usin DSRC (rate, range) adaptive schemes can outperform non-adaptive ones

Besides range/power control, we can also adapt rate based on estimation

error of d namical s stem and channel con estion

82

CEE: Systems

UCB

The VANET Communit


83/98

V2VCOM 2005 - 2008

Autonet 2006 - 2008

Wivec 2007 - 2008

a meeting

MOVE 2008

83

CEE: Systems

UCB


84/98

.

84


85/98

85


86/98

86


87/98

Patting ourselves on the back 1Decentralized Information A lications


88/98

Jerbi07

Accident report, parking,data disseminsation trafficmonitoring Okada 08, Jerbi07,

Sormani06, Caliskan 06,

Protocols Car torrent Gerla07 Multi-cast Ozguner04.

Lee08

88

CEE: Systems

UCB

Patting ourselves on the back 2Understandin the V2V Channel and


89/98

Channels Stancil07

Ramachandran07

Zhang08

Taliwal05

Antennas and Phyd iesign Bahai08 Malarky07,

89

CEE: Systems

UCB


90/98

Patting ourselves on the back 4Safet

250)


91/98

250

l ( m s ) Feasible

Sengupta04

Elbatt06 150 t i o n

I n t e r v

Labersteaux06

Hassan08 100 g e G e n e r a

Jiang08

Hartenstein08

50 M e s s Infeasi e

0

0 50 100 150 200 250

Number of Interfering Vehicles

91

CEE: Systems

UCB

FCW Broadcast RateFCW Broadcast RateAda tationAda tation


92/98

Motivation: balance the factorscontributing to the packet Inter-reception time (IRT) # consecutive packet losses: favors

Inter-broadcast interval: favors highbroadcast rates

,range, 100 Bytes payload

Examine different Broadcastintervals:

50, 100, 200, , 700 msec

Conjecture: There is an optimal broadcast interval that minimizes IRT


93/98

Rate Ada tationRezaei Phd thesis08Huang IEEE MSC08


94/98

94

CEE: Systems

UCB


95/98

Transit Si nal Priorit

Traffic Signal

Transit Vehicle

Transit Vehicle Stop

up to 1000 ft

Grass DividerCollisionAvoidance

E-Transaction: gas, movie, .

Gas Pumps

95

Not to Scale

IDB DataTransfer

* Graphic created from Broady Cash (ARINC)

DSRC as per FCC Ruling in 2004


96/98

DSRC has 75 MHz Spectrum (1 control channel, 6 service channels) DSRC supports the non-exclusive geographic area licensing

Each licensee can operate any of the service channels A license is for an area-of-operation (e.g., county, state)

Control channel is for safety messages and service announcements Three levels of priority: Safety of Life, Safety, Non-Safety

The service channels are used for non-safety related data traffic (e.g.

e-commerce, infotainment) ASTM standards committee voted in 2000 to base it on 802.11a

96

Now IEEE 802.11p

Typical DSRC Data Traffic Requirements

Application Packet Size(Bytes) AllowableLatency NetworkTraffic Comm. Range(m) Priority


97/98

(Bytes) /Bandwidth

Latency(ms)

TrafficType

(m)

Intersection CollisionWarning / Avoidance

~100 ~100 Event 50 300 Safety of Life

Forward Collision Warning ~100/ ~100 Periodic 50 300 Safety of ~10Kbps Life

Work Zone Warning ~100

~

~1000 Periodic 50 300 Safety

Transit Vehicle SignalPriority

~100 ~1000 Event 300 1000 Safety

-

Service Announcements ~100/~2Kbps

~500 Periodic 0 90 Non-Safety

97

Movie Download(2 hours of MPEG 1) : 10min. download time

> 20Mbps N/A N/A 0 90 N on-Safety

New approach needs new

Hi, Im transportation guy!

r en s


98/98

r en s,

your old

friend Car Company Hi, Im New. My nameIs Iphone

98

CEE: Systems

UCB

Raja Sengupta Wireless Networks Forcars

Documents