Smart Road Wireless Network for Intelligent Transportation System Development

1. Smart Road:Wireless Networks for Intelligent Transport system Kun-chan LanNICTA NCTU

2. About me • Graduated from USC in 2004 • Currently working as a researcher at National ICT Australia (NICTA) • Past research • Internet measurements, traffic modeling and simulations • Current research • Wireless mesh networks and vehicular ad-hoc networks NCTU

3. About NICTA • A national research institute funded by Australia Government • Our research staff includes regular full-time researchers and contributed staff from major universities such as Australian National Univ., Univ. of Sydney, Univ. of Melbourn, New South Wales of Univ. etc • Our focus • Research, commercialization, education, collaboration NCTU

4. About NICTA • A number of research labs • located in Sydney, Canberra and Melbourn • A variety of research programs • Empirical Software Engineering; • Interfaces, Machines, And Graphic ENvironments • Networks and Pervasive Computing. • Embedded, Real-Time, and Operating Systems • Formal Methods • Symbolic Machine Learning and Knowledge Acquisition • Statistical Machine Learning; • Systems Engineering and Complex Systems • Wireless Signal Processing • Logic and Computation; • Autonomous Systems and Sensing Technologies • Statistical Machine Learning. • Sensor Networks; • Network Information Processing. NCTU

5. What is Intelligent Transportation System (ITS) ? • Computer and communication technologies applied to management of transportation systems • To manage it in a safe and efficient manner • To monitor traffic conditions (accident, incidents, construction work, weather, major events) • Control traffic flow • To provide information to the traveling public about traffic conditions NCTU

6. Types of ITS implemented – infrastructure NCTU

7. Types of ITS implemented – vehicles NCTU

8. Why ITS? • Improved safety to drivers • e.g. reduce accident • Improved traffic efficiency • e.g. reduced traffic congestion • Improved environmental quality • e.g. reduced fuel/exhaust • Improved economic productivity • e.g. broadband service on buses NCTU

9. Improve safety • highway deaths > 40K in 2003 for US alone • Studies showed the use of ramp meters reduce accidents 15-50%. • In-vehicle computer visioning cameras • warn operators of drowsy driving behavior. • Results showed the system improved safety and decreased fuel consumption 15%. • Radar sensors on trucks • warn operators of obstacles in blind spots • at-fault accidents decreased 34% in 1 year. NCTU

10. Improve efficiency • in-vehicle navigation systems • a travel time savings of more than 10% • intelligent cruise control vehicles (ICC) • use road sensor data to optimize vehicle speeds and match signal timing • increase link capacity 3-6%. NCTU

11. Improve environmental quality • E-Zpass (an electronic toll collection) in NJ • saves: 1.2 mil gallons of fuel/yr, 0.35 tons of CO/day, and 0.056 tons NOx/day.  • Similar study from Baltimore • reduced hydrocarbons and Carbon monoxide emissions by 40-63%, and reduced emissions of Nitrogen oxides by approximately 16%. NCTU

12. ITS market • Promising market • US market for ITS is estimated to grow from $5 billion to $35 billion by 2010. • $700 billion is expected to be spent on transport infrastructure in the Asia Pacific market • In 2005, The Minister for ICT in Australia launched a new industry cluster in Victoria for the ITS market • the HK Government proposed to spend US$423 million on ITS in the next decade • In Japan, the annual market size has been estimated at 4 billion ECU by 2010. • The European standardized GSM-R • cellular solutions in the transportation sector • $5 billion new market in Europe within five years NCTU

13. Today’s talk • Part 1 • A brief talk about a project (STaR) we recently started at NICTA • A wireless mesh network for ITS • Not much results at this point, only architecture overview for today • Part 2 • two vehicular-network applications • MOBNET – A NEMO-based Network Mobility Testbed • MOVE – A Mobility mOdel generator for VANET • Only overview talk today, no discussion on math or protocol NCTU

14. About STaR (Smart Transports and Roads) • A multi-million research project we recently started at NICTA • Only a few months old • Collaborating with New South Wales Road and Traffic Authority (NSW RTA) • NSW RTA is the creator of SCATS, a real-time traffic management system • SCATS is used in Sydney and ~80 other cities around the world • It is expected that some outcome of STaR project can be integrated into SCATS in the future NCTU

15. SCATS (The Sydney Coordinated Adaptive Traffic System)) • An adaptive traffic control system • Goal • Minimize vehicle travel time when traffic is light • Maximize road capacities when traffic is heavy • Components of SCAT • Subsystem: 1-10 intersections • Local controller: one at each intersection • Critical intersection: need accurate timing control • Non-critical intersections synchronize with the critical intersection • Regional computer: control up to 64 subsystem Regional computer subsystem subsystem subsystem NCTU

16. SCATS (The Sydney Coordinated Adaptive Traffic System)) • Local controller: optimize local traffic flows • On a phase-by-phase basis • Phase length: time from one green to next green • Regional computer: optimize subsystem capacity • On a cycle-by-cycle basis • One cycle contain multiple phases: typically 40s-150s • All intersections in the same subsystem has the same cycle Regional computer subsystem NCTU

17. Goal of STaR • Improved traffic flow • Improved public safety • Improved performance, efficiency and running cost for public transports • Improved SCATS revenue in the multi-billion dollar ITS market NCTU

18. Benefit from working with RTA • Access to real-time and historical road traffic data • Access to public infrastructure (traffic controller,video camera, road-side sensors, etc) • Access to other RTA systems NCTU

19. Problems with the existing RTA network • Rely on a fixed communication infrastructure (ISDN and dial-up) • Costly to install, operate and maintain • Easy to be damaged • Low bandwidth (< 32KB/s): inflexible in its application NCTU

20. Wireless mesh networks for ITS • Replacement of current system is highly sought after • By RTA and other traffic authorities elsewhere • But commercial off-the-shelf systems (e.g. DSL, GPRS) don’t provide required reliability and timeliness, and still incur costs • NICTA proposes a multiple-hop wireless mesh network to replace the dial-up network • Easy deployment • Infrastructureless: lower maintenance cost • Initially, a 16-node testbed in Sydney CBD (Central Business District) area NCTU

21. Requirement for the test-bed • Representative locations • Foliages/trees • Pedestrians • Passing traffic • High-rise buildings • Easy access • Close to NICTA • Cover at least one critical intersection • Multiple paths available for each source/destination pairs • can provide external source of power to mesh router NCTU

22. Why wireless mesh networks? • Wireless mesh networking is a promising technology to replace current system • ease and speed of installation, without reliance on a telecommunications carrier or dependence on their time frames; • lower on-going costs than the current system, with no annual or monthly rental or service fees; • flexibility to connect new locations, for new intersections or during road works or emergencies; • Others also recognise that the potentials of wireless meshes NCTU

23. Commercial product • Meshnetworks Inc (now acquired by Motorola) • Tropos Networks • LocustWorld (MeshAP) • Intel • Nortel • Microsoft • Kiyon • Radiant Networks (Cambridge-based, work with BT) • Invisible Networks • Green Packet Inc. (M-Tapei) • SeattleWireless, NYCWireless,… • many others.. NCTU

24. Test-beds In academia • MIT (Roofnet project) • Rice university (TFA project) • Berkeley (DGP project @ India) • Trinity College @ Dublin (WAND project) • University of Massachusetts @ Ahmerst (Diesel Net) • UC Santa Barbara • UC San Diego • State University of New York @ Stony Brook • UCLA • others… NCTU

25. STaR network topology public transport Regional computer Traffic controller Mesh box camera road-side sensor Internet NCTU

26. Test-beds @ NICTA • We are currently building two test-beds at NICTA • Indoor – Linksys WRT54GS • ~$100 • Linux-based firmware • OpenWRT support • For research above transport layer • Outdoor – Soekris boards • ~$300 • Use Compact Flash card for OS • Customized MAC • Can use any type of radio • We’re only interested in research Issues above MAC NCTU

27. Current testbed activities • Wireless survey • Building mesh routers with soekris boards • Integrate mesh routers with SCATS simulator • Integrate mesh routers with real SCATS traffic controller • Network management NCTU

28. Wireless survey • Understand the radio property in real word • What to measure • Signal strength quality • Throughput • Packet losses • MAC layer re-transmission • as a function of • Distance • Transmission rate • Type/height of antenna • Number of MAC RETRY • Two phases • Open-space measurements • Intersection measurements NCTU

29. Multi-radio mesh router • Soekris net4521 • 133 Mhz AMD • 64 Mbyte SDRAM • Use CF card for OS (pebble linux) • 2 Ethernet ports • 1 Serial port, DB9. • 1 Mini-PCI slot • 2 PC-Card slot for wireless adapters • Board size 9.2" x 5.7" • External power supply 11-56V DC • Operating temperature 0-60 °C NCTU

30. Integrating with traffic simulator • Test mesh router on the traffic simulator before integrating with real traffic controller • A micro-traffic simulator Paramics is used, located at UNSW • One high speed link between UNSW and NICTA that allows the mesh router to talk to the Paramics simulator remotely NCTU

31. Integrating with SCATS traffic controllers • SCATS system • A large number of kerbside controllers that control traffic signal • A set of regional controllers that control kerbside controllers • Star topology (Masterlink mode) • Currently connected by leased lines via a Bell 103 modem at 300 baud • Some kerbside controllers have a special role for synchronizing signal timing when the regional controller is down (Flexilink mode) • Each kerbside controller has a 25-pin RS-232 connector for external access kerbside controller regional controller critical intersection NCTU

32. Network management • Graphic interface that shows • Wireless connectivity • Network topology • Link latency • Link throughput • Routing path • Router status NCTU

33. practical Issues for street deployment • Waterproof/weatherproof • Power source • Antenna placement • Vandalism • Passing traffic NCTU

34. Research Challenge • Scalability • Connecting numerous road-side devices to SCATS • Need to Integrate video cameras: High throughput, low jitter • Reliability • Mission-critical data (e.g. accident detection, traffic signal control, etc) • Requires timely routing that is robust against faults in nodes or links • Low latency • SCATS is a real-time traffic control system (< 1 sec) • Heterogeneity • Requires support for different radio types • e.g. incrementally deploy new radio technologies NCTU

35. Research focuses • New multi-radio multi-channel MAC • Scalability/reliability/latency • Multi-path routing • Reliability/latency • Fault detection and recovery • Reliability • Network management • Very difficult to physically to access the mesh nodes once they are deployed • Need to mechanism to node diagnostics, software upgrade, etc • Communication between vehicles and road-side devices NCTU

36. Current status of STaR • 4 months old, not much technical results yet • 6 months to deliver a pilot testbed that controls real traffic light • 14 researchers/students working on this project • International research collaboration(U. Cal @ Davis and U. Texas @ Arlington) • Communication with Australian startups • We’ve developed a couple of applications for ‘vehicle to road-side’ component though • MOBNET • Network mobility testbed (LANMAN 2005) • MOVE • Mobility model generator (poster in MOBICOM 2005) NCTU

37. Today’s talk • Part 1 • A brief talk about a project (STaR) we recently started at NICTA • Part 2 • two vehicular-network applications • MOBNET – A NEMO-based Network Mobility Testbed • MOVE – A Mobility model generator for VANET NCTU

38. Mobile Network • Providing broadband service for public transport passengers is becoming a popular ITS service • E.g. Connexion by Boeing • Mobile Network (MN):a network that can move and attach arbitrary points in the Internet • Mobile Network • On-board LAN • Mobile Router: • manage movement of MN and provide Internet access to MNNs • Mobile Network Node (MMN) • MMN: a node in the MN • Local Fixed Node (LFN) • Visiting Mobile Node (VMN) • Standardized protocol: NEMO • Extension of MIPv6 NCTU

39. MOBNET - Network mobility testbed • Ways to conduct network research • Simulation • Emulation • Real implementation • Physical layer models in wireless simulations are typically over-simplified • Need realistic testbed • Existing wireless testbed • CMU (ad-hoc routing) • TAP, Roofnet (mesh network) • ORBIT (generic testbed) • Existing work does not support testing of network mobility protocols • Our contribution: a testbed for network mobility research MOBNET: The Design and Implementation of a Network Mobility Testbed", Kun-chan Lan, Eranga Perera, Henrik Petander, Christoph Dwertmann, Lavy Libman, Mahbub Hassan, IEEE LANMAN 05’ NCTU

40. NCTU

41. Mobile Nework testbed functionality • Emulation of a mobile network • Experimental control • Topology control • Mobility control • Management of the testbed NCTU

42. NCTU

43. Emulation of a Mobile Network • A NEMO-based mobile router • Extended from HUT MIPv6 • Built on Linux 2.4.26 • Support NEMO implicit mode • Can use link layer information to trigger handoff • Support Route Optimization • Extension of MIPv6 RO NCTU

44. Experimental control • Topology control • NIS NET network emulator • Mobility control • Mobility emulator (MobE) • Emulate the movement of MR • Emulate the variations of radio propagation by changing the transmission power of the AP • Can be driven used pre-made mobility patterns • Markov Chains model • For controllable experiments • Signal strength traces NCTU

45. Mobility emulator • Architecture • Input parser • Graphical interface • AP power level controller • AP power level control • web interface • telnet • Modeling signal strength variations • Discrete markov chain • e.g. frequency of changing from one power level to another, etc. NCTU

46. MR moving from ap1 toward ap2 Mobility Emulator interface NCTU

47. Testbed management • We’d like to make our testbed available to other researchers in the future • Remote management server • Remote users access • Testbed monitoring • Testbed maintenance NCTU

48. Node generator - Virtual MMNs • Currently each node is implemented via one single machine • Not scalable: how to emulate a large number of MMNs • Implement each MMN as a single process • Virtual interface: a single wireless interface is abstracted into multiple virtual interfaces • Each MMN connects to MR via a virtual interface NCTU

49. Effect of NEMO handoff on TCP and UDP • Experiment setup • 1 home network and 2 foreign networks • Traffic is sent from CN to MNN • Using Iperf to generate TCP and UDP traffic • UDP: 200Bytes at 100Kbits/s • Packet of a smaller size is more sensitive to the effect of handoff latency NCTU

50. Effect of NEMO handoff on TCP and UDP Use measurements on the Home network as a base line NCTU