1. MobiSteer: Using Steerable Beam Directional Antenna for Vehicular Network Access Vishnu Navda, Anand Prabhu Subramanian, Kannon Dhanasekaran, Andreas Timm-Giel, & Samir R. Das Originally Presented at MobiSys ‘07 Reviewed by Lauren Cohen on 2/12/08

2. Motivation Wireless communication between moving vehicles and roadside access points Three application types ◦ Traffic safety and information ◦ Mobile sensors ◦ Internet access for vehicle occupants Current connectivity is poor ◦ Access point inter-arrival time >> median connection time ◦ Link layer delivery rate ~80%

3. Design Goals Directional antennas to increase gain and reduce interference Steerable to maintain best link quality over longest duration Optimize handoff between access points

4. Omnidirectional: Directional:

5. Assumptions APs use omnidirectional antennas Vehicles access stationary network using one-hop links ◦ …in other words, we’ll be considering applications #2 (mobile sensors) and #3 (Internet access)

6. MobiSteer Architecture Overview

7. Hardware/Software Setup Multi-beam 2.4 GHz antenna ◦ One omnidirectional beam ◦ 16 45 ⁰ directional beams Computer-controlled steering commands via serial Data on captured packets logged to RF signature database during idle times

8. Operational Modes Cached ◦ Familiar territory ◦ RF signature database used to determine optimal steering and AP selection ◦ Databases can be downloaded from a server Online ◦ Previously untravelled routes ◦ Scans all beams/channels and determines best based in SNR value

9. Data Collection Link quality of received frames stored in RF signature database Passive scanning ◦ Monitors each beam/channel for any frame Active probing ◦ Periodic probe requests sent ◦ Responses from APs recorded ◦ Allows quicker sampling

10. Cached Mode Operation

11. Optimal AP and Beam Selection Computes best AP and beam for every point in trajectory ◦ Segments of length ∆ ◦ RF signature database queried for SNR Still need to deal with handoff latency

12. Optimal Handoff Algorithm Estimate speed of vehicle from RF signature database to calculate handoff latency between possible APs Use dynamic programming to select best APs to minimize latency between segments

13. Experimental Scenarios Controlled scenarios ◦ Single AP in empty parking lot ◦ Multiple APs in apartment complex  Typically two within hearing range  All on same channel In situ scenario ◦ Existing APs along campus roadways  Again all on same channel ◦ No actual data transferred

14. Experimental Results Controlled scenario ◦ Data collected on # of packets received and rate of transmission ◦ MobiSteer greatly increased both In situ scenario ◦ MobiSteer improved average SNR and distance from which each beam could be heard ◦ Alas, no actual data rate results

15. Online Mode Operation

16. Intricacies Constantly uses active probing Simple heuristic used to choose best AP/beam combination from probed data Only steering considered, as handoff and AP selection covered in other literature

17. Experimental Results Used controlled scenario setup Again measured # of packets received and data rate MobiSteer again improved both measurements over using an omnidirectional beam in online mode

18. Conclusions MobiSteer provides good alternative to omnidirectional beams for vehicular networking by improving connectivity duration and data rate Cached mode is superior to online mode Future ideas ◦ Use for localization of roadside APs ◦ Interface with cellular modem networks for additional connectivity

19. Questions What about communications between moving vehicles? Can this be used to improve cellular networks as well? How do we alleviate the overhead of active probing (since it’s necessary to build the RF signature database)?