Vehicular Network Applications VoIP Web Cab scheduling Congestion detection Vehicle platooning Road hazard warning Collision alert Stoplight assistant Toll collection Deceleration warning Emergency vehicle warning Border clearance Traction updates Flat tire warning Merge assistance

Congestion Detection Vehicles detect congestion when: # Vehicles > Threshold 1 Speed < Threshold 2 Relay congestion information Hop-by-hop message forwarding Other vehicles can choose alternate routes

Deceleration Warning Prevent pile-ups when a vehicle decelerates rapidly

Wireless Technologies for Vehicular Networks Cellular networks High coverage, low bandwidth, expensive WiFi networks Moderate coverage, high bandwidth, free Combine all of them to achieve low cost, high bandwidth, and high coverage

Aruna Balasubramanian, Ratul Mahajan Arun Venkataramani, Brian N Levine, John Zahorjan Interactive WiFi Connectivity from Moving Vehicles University of Massachusetts Amherst Microsoft Research University of Washington

Target Scenarios A car is within the range of multiple APs How common? Low data rate but low delay Alternatives?

Overview Internet Given enough coverage, can WiFi technology be used to access mainstream applications from vehicles? 7 Existing work shows the feasibility of WiFi access at vehicular speeds focus on non-interactive applications. e.g., road monitoring

Outline Can popular applications be supported using vehicular WiFi today? Performance is poor due to frequent disruptions How can we improve application performance? ViFi, a new handoff protocol that significantly reduces disruptions Does ViFi really improve application performance? VoIP, short TCP transfers 8

VanLAN: Vehicular Testbed Uses MS campus vans Base stations(BSes) are deployed on roadside buildings Currently 2 vans, 11 BSes 9

Measurement study Study application performance in vehicular WiFi setting Focus on basic connectivity Study performance of different handoff policies Trace-driven analysis Nodes send periodic packets and log receptions 10

Handoff policies studied Practical hard handoff Associate with one BS Current Ideal hard handoff Use future knowledge Impractical 11

Handoff policies studied Practical hard handoff Associate with one BS Current Ideal hard handoff Use future knowledge Impractical Ideal soft handoff Use all BSes in range Performance upper bound 12

Comparison of handoff policies Practical hard handoffIdeal hard handoffIdeal soft handoff Disruption Summary Performance of interactive applications poor when using existing handoff policies Soft handoff policy can decrease disruptions and improve performance of interactive applications 13

Outline Can popular applications be accessed using vehicular WiFi? How can we improve application performance? ViFi, a practical diversity-based handoff protocol Does ViFi really improve application performance? VoIP, short TCP transfers 14

Design a practical soft handoff policy Goal: Leverage multiple BSes in range How often do we have multiple BSes? Not straightforward Internet 15 Constraints in Vehicular WiFi 1. Inter-BS backplane often bandwidth-constrained 2. Interactive applications require timely delivery 3. Fine-grained scheduling of packets difficult

Why are existing solutions inadequate? Opportunistic protocols for WiFi mesh (ExOR, MORE) Uses batching: Not suitable for interactive applications Path diversity protocols for enterprise WLANs (Divert ) Assumes BSes are connected through a high speed back plane Soft handoff protocols for cellular (CDMA-based) Packet scheduling at fine time scales Signals can be combined 16

ViFi protocol set up Internet A B D C Vehicle chooses anchor BS Anchor responsible for vehicle’s packets Vehicle chooses a set of BSes in range to be auxiliaries e.g., B, C and D can be chosen as auxiliaries ViFi leverages packets overheard by the auxiliary 17

ViFi protocol (1)Source transmits a packet (2)If destination receives, it transmits an ack (3)If auxiliary overhears packet but not ack, it probabilistically relays to destination (4)If destination received relay, it transmits an ack (5)If no ack within retransmission interval, source retransmits A B D C A B D C Downstream: Anchor to vehicle Upstream: Vehicle to anchor 18 Dest Source Dest Source

Why relaying is effective? 19

Why relaying is effective? Losses are bursty Independence Losses from different senders independent Losses at different receivers independent AB D C Upstream 20 AB D C Downstream

Guidelines for probability computation 1. Make a collective relaying decision and limit the total number of relays 2. Give preference to auxiliary with good connectivity with destination How to make a collective decision without per-packet coordination overhead? 21

Determine the relaying probability Goal: Compute relaying probability R B of auxiliary B Step 1: The probability that auxiliary B is considering relaying C B = P(B heard the packet). P(B did not hear ack) Step 2: The expected number of relays by B is E(B) = C B ¢ R B Step 3: Formulate ViFi probability equation,  E(x) = 1 to solve uniquely, set R B proportional to P(destination hears B) Step 4: B estimates P(auxiliary considering relaying) and P(destination heard auxiliary) for each auxiliary 22 ViFi: Practical soft handoff protocol uses probabilistic relaying for coordination without per-packet coordination cost

ViFi Implementation Implemented ViFi in windows operating system Use broadcast transmission at the MAC layer No rate adaptation Deployed ViFi on VanLAN BSes and vehicles 23

Outline Can popular applications be accessed using vehicular WiFi? Due to frequent disruptions, performance is poor How can we improve application performance? ViFi, a practical diversity-based soft handoff protocol Does ViFi really improve application performance? 24

Evaluation Evaluation based on VanLAN deployment ViFi reduces disruptions ViFi improves application performance ViFi’s probabilistic relaying is efficient Also in the paper: Trace-driven evaluation on DieselNet testbed at UMass, Amherst Results qualitatively consistent 25

ViFi reduces disruptions in our deployment Practical hard handoff ViFi 26

ViFi improves VoIP performance > 100% Disruption = When mean opinion score (mos) is lower than a threshold Length of voice call before disruption Use G.729 codec 27 seconds ViFi Practical hard handoff

ViFi improves performance of short TCP transfers Median transfer time (sec) > 50% > 100% Number of transfers before disruption Workload: repeatedly download/upload 10KB files Disruption = lack of progress for 10 seconds 28 ViFi Practical hard handoff

ViFi uses medium efficiently efficiency Efficiency: Number of unique packets delivered/ Number of packets sent It’s efficient for their testbed, but may not be the case in general. Why? 29 ViFi Practical hard handoff

Conclusions Improves performance of interactive applications for vehicular WiFi networks Interactive applications perform poorly in vehicular settings due to frequent disruptions ViFi, a diversity-based handoff protocol significantly reduces disruptions Experiments on VanLAN shows that ViFi significantly improves performance of VoIP and short TCP transfers 30

Comments Interesting problem domain Target low-bandwidth applications, for which cellular networks are sufficient Have multiple APs within range tuned into the same channel May not be common and lose spatial diversity Use the lowest data rate Common to have multiple or fewer than 1 relay(s) for each tx Relay is not compelling Uplink: sufficient to relay data to one AP Downlink: if best AP is selected, the need for relay is low If relay has to be used, MORE like opportunistic routing may be more efficient They dismissed opportunistic routing due to its potential large delay due to batch But their delay can be high since retx timeout is generally large in order to account for variable contention delay

32

People want to communicate while on the move Average one way commute (2005): US: 24.3min, World: 40min Passengers want to watch videos, listen to songs, etc. Why not just use cellular networks? Expensive: $30-$60/month 5GB/month -> 2Kbps! 40% 3G capable devices have no 3G plan iPod Touch sales ~ iPhone sales Bandwidth and backhaul limitations Limited video quality (96-128kbps, < 10min long) Carriers interested in WiFi offloading Arms race between Increase in cellular bandwidth Higher resolution screens and videos Goal: Enable high bandwidth applications (e.g., video) in vehicular networks via WiFi Motivation 33 ✗ ✔

34 Opportunistic WiFi connectivity Interne t Compelling usage scenario Taxi companies provide value-added services to passengers Previous work: low-bandwidth applications We focus on delivering high-bandwidth content e.g. video streaming Devices in vehicles contact roadside APs Gas stations and local shops deploy APs Passengers watch videos, download files

35 Synergy among connections High b/w, short-lived High b/w, high delay Low b/w, persistent High b/w, low coverage VCD High b/w, persistent

36 Contributions New techniques for replication optimization Goal: Fully utilize wireless bandwidth during contact Optimized wireline replication to Internet-connected APs Replication using vehicular relays to unconnected APs Use mesh for replication and caching New algorithm for mobility prediction Predict set of APs that will be visited by vehicle Critical for success of replication techniques Algorithm: voting among K nearest trajectories

Trace-driven simulation and emulation San Francisco cabs, Seattle buses, Shanghai cabs Two testbeds on UT campus b: 14 APs deployed inside 8 campus buildings, 20-60ft from the road n: 4 APs outdoor, 1-5ft from the road Smartphone and laptop clients HP iPAQ and HTC Tilt Stream H.264 videos at 64Kbps 37 Evaluation Methodology ,8,9,10, 11,12 13, 14

Summary: Vehicular Content Distribution KNT: A new mobility prediction algorithm Based on voting among K nearest trajectories 25-94% more accurate than 1 st and 2 nd order Markov models A series of novel replication schemes Optimized wireline replication and mesh replication Opportunistic vehicular relay based replication Extensive evaluation: simulation + testbed + emulation Simulation using San Francisco taxi and Seattle bus traces 3-6x of no replication, 2-4x of wireline or vehicular alone Full-fledged prototype deployed on two real testbeds 14-node b testbed and 4-node n testbed x gain over no replication Emulab emulation with real AP/controller and emulated vehicles Show system works at scale and is efficient Validate our trace-driven simulator