1
V2X on Phones: Enabling next-generation mobile ITS apps Jason H. Gao [email protected] Next%gen ITS Apps via D2D V2X & D2D Radios on Phones RoadRu nner (Gao et. al., 2014) use s V2X commun ication s to r ealize city4 wide fine4grained traffic management. Roads or ge ograph ic areas are d ivide d into r egion s, and some can be designate d a s c ongest ion c ontro lle d. To enter such a region, a vehicle mu st poss es a d igital t oken. When it exit s, it passe s the token to an other vehic le v ia V2 V, or back to a fal lback s erver if the re are no nearby vehicles that need it. Li,Shiuan Peh [email protected] RoadRunner Automotive connectivity standards promise to usher in new apps and services that utilize vehicle4to4vehicle communications, and mobile app trends point to the potential of direct device4to4device connections. The convergence of the two, made possible by new hardware, system software, and programming abstractions, promises to realize next4generation mobile ITS apps. The city4scale impact of such convergence can now be projected using new evaluation infrastructure that captures interactions between humans, vehicles, devices and networks. Evaluation & Test Infrastructure V2X & D2D System Software SignalGuru (Koukoum idi s et. al., 2011) ru ns on win dsh ield4m ounte d iPho nes that collaborat ively sen se, share, and learn traffic signal transitions. It prov ides Green Light Opt imal Speed Advisory (GLOSA) to drivers ,re ducing fuel cons umpti on by 20%. The p hone s comm unicated via ad4hoc WiFi , wh ic h has every short ra nge , nec essitat ing a r oad4s ide relay node. Longer4range V2 V wou ld elim inate the need for such infrastructure. SignalGuru Gao et. al. (2012) impleme nted a clon e of Panoram io (a pop ular location4 based photo sharing app) that leverages D2D. A geogra phical ar ea is divided into a gr id of reg ions, and the pho nes wit hin each region collaborat ively h ost locatio n4base d se rvices. Apps can access data in oth er regions via m ult i4hop netwo rking bet ween reg ion s. The D2D4 enabled app impr oved latency 10X versu s a 3G4on ly ba sel ine a nd 2 X vers us 4G4only,and offloaded cellular ne twork accesse s by 9 6%. A dep loyment with 20 device s was limited to re latively sma ll regio ns and d istance s betwee n phone s, again due to the limited range of ad4h oc WiFi. Longer4range a nd faste r V2X wou ld a llo w much larger reg ion s, and improve service responsiveness. Panoramio0clone DIPLOMA(DIstributed Programming LayerOver Mobile Agents) isa programming layer and distributed share d memory sy stem that provide s cohere nt relaxed4con sistency accessto data re sid ing on different mo bile ph one sacrossa large geograph ic area that is divide d into regi ons. It abstract s a collection of unr elia ble mob ile no des int o a stationa ry reliab le “virtua l computer core”, wh ile a ddres si ng practical i ssue s ari sing from unpre dic tabl e mob il ity and unre liab le net worki ng by jud ic iou sly usi ng the c loud for occasional coordination. DIPLOMA MobiStream s is a Distributed Stream Processing System (DSPS) that ru ns direct ly on smartpho nes int erconnected by D2D via ad4 hoc WiFi, wherea s other DSPSsty pically run on cl uster s of re lia ble se rvers interco nnected by high4 spee d Et hernet. To overcome the unrel iabi lity of us ing smar tpho nes , Wa ng et. al. (201 2) bu ild up a re liab le DSPS u sing check4pointing, with two new approaches to reduce overhead: 1) Token:tr iggered checkp oint ing: sour ce operator s can gener ate token s t hat f low down and t rigger che ckpoint ing at ea ch do wnstream operator , avoid ing the red undant data saves that occur in prior fault tolerance schemes. 2) Bro adcast: based checkpo int ing: che ckpoint dat a tr ansm iss ion i s spl it into mult iple phases, and u ses un reliable UDP broa dca sts in the first p hase s t o avoid red undant transmission. MobiStreams Bus Capacity Pred iction (BC P) (Wa ng et. al., 2012) isan app that pre dicts the numb er of passenge rs on inc omi ng b use s so t hat u sers at bus st ops c an make dec isi ons l ike wa iting for a less cro wded bus or boar ding a lternative buse s. Wang et. al. dep loyed it alo ng a campus bu s route in Singapore , on 8 ph ones , an d use s stat ist ic al mode ls for (dis)embark ing passengers, collected via o n4vehicle infrare d sen sors an d cameras installed at bus stops. Bus Capacity Prediction RoadRunner prototype mounted in vehicle Screenshot of SignalGuru app The DSPS for each bus stop in the BCP app. Operators with the same color run on the same node. 802.11p is a stan dard for Vehicle% 2%Vehicle (V2V) and Vehicle%2%Anything (V2X) wireless comm unications. It is sim ilar t o 802.11a , but t o realize 8 02.11p rad ios wit hin the power an d si ze constra ints of p hone s, C hoi et. al. (2 014) leverage the LEES pr ocess to de sig n a suita ble RF front4e nd wh ich can be integrated o n pho nes to rea lize Dev ice% to%Device (D2D) radios. Example of 802.11p front%end integration on mobile device The apps de scribed have bee n limited to small pilot deployments of several vehicle s. Large4scale simulat ion infrastr ucture can enable eva luation of city4scale im plications of V2X4enabled ITS apps. SimMobility SimMo bi lity (B asak et. a l., 2013) is an integrated simulat ion p latform that enab les developmen t a nd evaluat ion of fut ure ITS scena rio s, with integrated mo dels of human and commercia l activ itie s, la nd use , trans portat ion ,e nvironm ental im pacts, and en ergy use. As an example of t he potent ial of large4scale evaluat ion infrastr ucture like Sim Mob ility to measure city4scale impact, we were able to simulate 74 ,904 vehicle trips over 24 hours of simu lated t ime for Road Runn er, compare d to 10 veh icles fo r 2 hour s in the deployment. Similitude Simi lit ude (Hetu et. a l., 2 014) comb ine s Sim Mob ili ty w ith emulate d Andro id devices and th e n s43 netw ork simu lator to all ow sim ulat ion of m obi le ITS a pps at city4scale. The Andr oid a pps requ ire s ome ru dimenta ry mod ification s, but S imi litu de prov ides a Java library, s immob ili ty4andro id ,t o ea se integrati on and provi de sh ims for f unctio nality l ike location updates, timing, and message passing.

V2X on Phones...evaluation infrastructure that captures interactions between humans, vehicles,devices andnetworks. Evaluation(&(Test(Infrastructure V2X(&(D2D(System(Software SignalGuru(Koukoumidis

  • Upload
    others

  • View
    4

  • Download
    0

Embed Size (px)

Citation preview

Page 1: V2X on Phones...evaluation infrastructure that captures interactions between humans, vehicles,devices andnetworks. Evaluation(&(Test(Infrastructure V2X(&(D2D(System(Software SignalGuru(Koukoumidis

V2X on Phones:Enabling next-generation mobile ITS apps

Jason&H.&[email protected]

Next%gen(ITS(Apps(via(D2D

V2X(&(D2D(Radios(on(Phones

RoadRunner (Gao et. a l. , 2014) use sV2X commun ication s to r eali ze city4wide fin e4grained traffic management.Roads or ge ograph ic areas are d ivide dinto r egion s, and some can bedesignate da s congest ion contro lle d. Toenter such a region , a vehicle mustposses a d igital t oken. When it exit s, itpasse s the token to another vehic le v iaV2 V, or back to a fal lback server if the reare no nearby vehicles that need it.

Li,Shiuan&[email protected]

RoadRunner

Automotive connectivity standards promise to usher in new appsand services that utilize vehicle4to4vehicle communications, andmobile app trends point to the potential of direct device4to4deviceconnections. The convergence of the two, made possible by newhardware, system software, and programming abstractions,promises to realize next4generation mobile ITS apps. The city4scaleimpact of such convergence can now be projected using newevaluation infrastructure that captures interactions betweenhumans, vehicles, devices and networks.

Evaluation(&(Test(Infrastructure

V2X(&(D2D(System(Software

SignalGuru (Koukoum idiset. al. , 2011) runson windsh ield4m ounte d iPhones thatcollaborat ively sen se, share, and learntraffic signal trans itions. It prov ides GreenLight Opt imal Speed Advisory (GLOSA) todrivers ,re ducing fuel consumptionby 20%.The phone s comm unicated via ad4hocWiFi , wh ich has every short ra nge ,necessitat ing a r oad4s ide relay node.Longer4range V2V wou ld elim inate theneed for such infrastructure.

SignalGuru

Gao et. al. (2012) impleme nted a clone of Panoram io (a popular location4 based photosharing app) that leverages D2D. A geogra phical ar ea is divided into a gr id of reg ion s,and the phones wit hin each region col laborat ively host location4base d se rvices. Appscan access data in other regions via m ult i4hop netwo rking between reg ion s. The D2D4enabled app impr oved latency 10X versu sa 3G4on ly ba sel ine a nd2 X versus 4G4on ly,andoffloaded cel lular ne twork accesse s by 9 6%. A dep loyment with 20 device s was l imitedto re latively sma ll regions and d istance s betwee n phone s, again due to the l imitedrange of ad4hoc WiFi. Longer4range a nd faste r V2X wou ld a llow much larger reg ion s,and improve service responsiveness.

Panoramio0cloneDIPLOMA(DIstributed Pr ogramming LayerOver Mobi le Agen ts) i sa p rogramming layerand distributed share d memory sy stem that provide s cohere nt relaxed4con sistencyaccessto data re sid ing on different mobile phone sacrossa large geograph ic area that isdivide d into regions. It abstract s a col lection of unr elia ble mob ile nodes int o astationa ry reliab le “virtua l computer core”, wh ile a ddressing practical i ssue s arisingfrom unpre dictable mob il ity and unre liab le networking by jud iciou sly using the c loudfor occasional coordination.

DIPLOMA

MobiStream s is a Distributed Stream Process ing System (DSPS) that runs direct ly onsmartphones int erconnected by D2D via ad4 hoc WiFi , wherea s other DSPSsty pically runon cluster s of re lia ble se rvers interconnected by high4 spee dEt hernet. To overcome theunrel iabi lity of us ing smar tphones ,Wang et. al. (201 2) bu ild up a re liab le DSPS u singcheck4pointing, with two new approaches to reduce overhead:1) Token:tr iggered checkp oint ing: sour ce operator s can gener ate token s t hat f low

down and t rigger che ckpoint ing at ea ch downstream operator , avoid ing the red undantdata saves that occur in prior fault tolerance schemes.2) Bro adcast: based checkpo int ing: che ckpoint dat a tr ansm iss ion is spl it into mult iple

phases , and u ses un reliable UDP broa dca sts in the first p hase s t o avoid red undanttransmission.

MobiStreams

Bus Capacity Pred iction (BCP) (Wang et. al. , 2012) i san app that pre dicts the number ofpassenge rs on incoming buse s so t hat u sers at bus st ops canmake dec isions l ikewa itingfor a less crowded bus or boar ding a lternative buse s. Wang et. al. dep loyed it along acampus bu s route in Singapore , on 8 phones , and use s stat ist ical mode ls for(dis)embark ing passengers , collected via on4vehicle infrare d sen sors and camerasinstalled at bus stops.

Bus<Capacity<Prediction

RoadRunner( prototype( mounted( in(vehicle(

Screenshot( of(SignalGuru app

The(DSPS(for(each(bus(stop( in( the(BCP( app.Operators( with( the(same( color( run(on( the(same( node.(

802.11p is a standard for Vehicle% 2%Vehicle (V2V) and Vehicle%2%Anyt hing (V2X)wireless comm unications. It is sim ilar t o 802.11a , but t o real ize 8 02.11p rad ios wit hinthe power and size constra ints of phone s, Choi et. al. (2 014) leverage the LEES pr ocesstode sig na suita ble RF front4e nd wh ich can be integrated on phones to rea lize Dev ice%to%Device (D2D) radios.

Example(of(802.11p( front%end( integration( on(mobile( device

The app s de scribed have bee n l imited to small pi lot d eployments of several vehicle s.Large4scale s imulat ion infrastr ucture can enable eva luation of city4scale im plications ofV2X4enabled ITS apps.

SimMobilitySimMobility (Basak et. a l. , 2013) is an integrated s imulat ion p latform that enab lesdevelopmen t a nd evaluat ion of fut ure ITS scena rio s, with integrated models of humanandcommercia lactiv itie s, la nd use , transportat ion ,e nvironm ental im pacts, and energyuse.

As an example of t he potent ial of large4scale evaluat ion infrastr ucture l ike SimMob ili tyto measu re city4scale impact , we wer e ab le to s imulate 74 ,904 vehicle trips over 24hours of s imu lated t ime for RoadRunner, compare d to 10 veh icles fo r 2 hour s in thedeployment.

SimilitudeSimilit ude (Hetu et. a l. , 2 014) comb ine s SimMob ili ty w ith emulate d Andro id devicesand the n s43 network simu lator to allow sim ulat ion of m obi le ITS a pps at city4scale. TheAndr oid a pps requ ire some rudimenta ry mod ification s, but S imi litu de prov ides a Javalibrary, s immob ili ty4andro id ,t oea se integration and provide sh ims for f unctionality l ikelocation updates, timing, and message passing.