Towards Automatic Phone-to-Phone Communication for Vehicular - PowerPoint PPT Presentation

Towards Automatic Phone-to-Phone Communication for Vehicular Networking Applications Shaohan Hu , Hengchang Liu, Lu Su, Hongyan Wang,Tarek Abdelzaher, Pan Hui, Wei Zheng, Zhiheng Xie, John Stankovic

Motivation

Motivation richer traffic info infotainment opportunities Gas Coupon! higher road awareness … and more!

Goal Build a Phone-to-Phone communication system for the tens of millions of vehicles on the roads today WiFi AP Internet WiFi AP

Goal Build a Phone-to-Phone communication system for the tens of millions of vehicles on the roads today WiFi AP Internet Key requirements WiFi No change to the existing AP infrastructure / protocols Transparent to the end-users (e.g., no rooting / jailbreaking)

System Model

System Model Hotspot Client Infrastructure Mode Mode (WiFi AP) Vehicle-resident phones toggle between hotspot and client modes

System Model Hotspot Client Infrastructure Mode Mode (WiFi AP) Vehicle-resident phones toggle between hotspot and client modes Only considering pairwise communication: T-Drive dataset => ~80% encounters are pairwise (T-Drive: ~10k taxicabs’ 1-week traces in Beijing, collected by MSRA)

Problem Statement How to design the toggling scheme for optimal System Efficiency?

Problem Statement How to design the toggling scheme for optimal System Efficiency? T data transfer can take place T phones in range with each other (or AP)

Problem Statement How to design the toggling scheme for optimal System Efficiency? T data transfer can take place T phones in range with each other (or AP) When to For how long do How often do toggle Phone-Phone (or Phone-Phone (or Phone-AP) meet? Phone-AP) meet? modes?

Analytical Formulation

Analytical Formulation h r ∗ , s ∗ i = arg max r,s E M 1 ,M 2 [ E ( β ) T 1 + E ( γ ) T 2 ]

Analytical Formulation h r ∗ , s ∗ i = arg max r,s E M 1 ,M 2 [ E ( β ) T 1 + E ( γ ) T 2 ] : Phone-Phone (or Phone-AP) meeting duration M 1 (or M 2 ) : Phone-Phone (or Phone-AP) expected T 1 (or T 2 ) transmission duration : Phone-Phone (or Phone-AP) meeting rate β (or γ )

Solution Sketch ( meeting times and rates ): M 1 , M 2 , β , γ estimated from empirical data ( expected transmission times ): T 1 , T 2 ? derived analytically ( optimal mode-toggling policy ): h r ∗ , s ∗ i solved using off-the-shelf non-linear optimization solver

Expected Transmission Time Phone-to-Phone

Expected Transmission Time Phone-to-Phone Define a periodic function  => during mode switching 0 , 0 ≤ t ≤ t 0   => in hotspot mode 1 , t 0 < t ≤ t 0 + r  f ( t ) = => during mode switching 0 , t 0 + r < t ≤ 2 t 0 + r   − 1 , 2 t 0 + r < t ≤ 2 t 0 + r + s => in client mode  with period . f = 2 t 0 + r + s

Expected Transmission Time Phone-to-Phone Define a periodic function  => during mode switching 0 , 0 ≤ t ≤ t 0   => in hotspot mode 1 , t 0 < t ≤ t 0 + r  f ( t ) = => during mode switching 0 , t 0 + r < t ≤ 2 t 0 + r   − 1 , 2 t 0 + r < t ≤ 2 t 0 + r + s => in client mode  with period . f = 2 t 0 + r + s Then, the time since meeting when two phones establish connection is t ∗ = min t { t : f 1 ( t ) f 2 ( t ) < 0 }

Expected Transmission Time Phone-to-Phone Define a periodic function  => during mode switching 0 , 0 ≤ t ≤ t 0   => in hotspot mode 1 , t 0 < t ≤ t 0 + r  f ( t ) = => during mode switching 0 , t 0 + r < t ≤ 2 t 0 + r   − 1 , 2 t 0 + r < t ≤ 2 t 0 + r + s => in client mode  with period . f = 2 t 0 + r + s Then, the time since meeting when two phones establish connection is t ∗ = min t { t : f 1 ( t ) f 2 ( t ) < 0 } Therefore, the expected phone-to-phone transmission time is T 1 = E t 1 ,t 2 [ M 1 − t ∗ ] = 1 Z Z ( M 1 − t ∗ ) dt 2 dt 1 f 2 t 1 t 2

Expected Transmission Time Phone-to-Phone E ( M 1 − t ∗ ) t 1 t 2 t ∗ M 1 [0 , t 0 ) [0 , t 0 ) 0 ∞ ∼ 3 ( M 1 − r ) 3 + 1 ( M 1 − t 0 ) 2 − 1 1 ⇥ 3 ( M 1 − t 0 − r ) 3 ⇤ [0 , t 0 ) [ t 0 , t 0 + r ) M 1 ≥ t 0 + r 2 f 2 1 ( M 1 − t 0 ) 2 t 0 − 1 ⇥ 3 ( M 1 − r ) 3 ⇤ or 2 t 0 + r − t 2 r ≤ M 1 < t 0 + r 2 f 2 t 0 2 f 2 ( M 1 − t 0 ) 2 [ t 0 , t 0 + r ) [0 , t 0 ) t 0 ≤ M 1 < r ⇣ ⌘ M 1 2 0 − 1 2 t 2 3 t 3 M 1 ≥ t 0 f 2 [0 , t 0 ) [ t 0 + r, 2 t 0 + r ) max( t 0 − t 1 , 2 t 0 + r − t 2 ) 0 1 3 f 2 M 3 M 1 < t 0 1 ⇥ 1 1 0 − 1 3 t 3 2 ( M 1 + s ) t 2 ⇤ 0 + M 1 st 0 M 1 ≥ t 0 f 2 [0 , t 0 ) [2 t 0 + r, f ) t 0 − t 1 � 1 1 2 sM 2 1 − 1 6 M 3 � M 1 < t 0 f 2 1 h 6 ( M 1 − r ) 3 i 6 ( M 1 − t 0 ) 3 + 1 ( M 1 − t 0 ) 2 − 1 r − t 0 2 M 1 ≥ r f 2 [ t 0 , t 0 + r ) [ t 0 , t 0 + r ) 2 t 0 + r − max( t 1 , t 2 ) 2 ( r − t 0 )( M 1 − t 0 ) 2 − 1 1 ⇥ 3 ( M 1 − t 0 ) 3 ⇤ t 0 ≤ M 1 < r f 2 h 6 ( M 1 − t 0 ) 3 i 1 − r − t 0 ( M 1 − t 0 ) 2 − 1 r 1 1 + 1 2 M 2 6 M 3 M 1 ≥ t 0 [ t 0 , t 0 + r ) [ t 0 + r, 2 t 0 + r ) 2 t 0 + r − t 2 f 2 2 � r 1 2 M 2 1 − 1 6 M 3 � M 1 < t 0 f 2 1 M 1 rs [ t 0 , t 0 + r ) [2 t 0 + r, f ) 0 ∼ f 2 Case analysis for expected transmission time

Expected Transmission Time Phone-to-Phone Collecting the cases: 2 rs = f 2 M 1 + I ( M 1 <t 0 ) f 1 ( r, s, M 1 ) + I ( M 1 ≥ t 0 ) f 2 ( r, s, M 1 ) + I ( t 0 ≤ M 1 <r + t 0 ) f 3 ( r, s, M 1 ) T 1 + I ( M 1 ≥ r + t 0 ) f 4 ( r, s, M 1 ) + I ( t 0 ≤ M 1 <s + t 0 ) f 5 ( r, s, M 1 ) + I ( M 1 ≥ s + t 0 ) f 6 ( r, s, M 1 ) , where the I ’s are indicator functions, and f ’s are s + r f 2 M 2 f 1 ( r, s, M 1 ) = 1 1  2 1 − 2 3( M 1 − t 0 ) 3 − 2 t 0 ( M 1 − t 0 ) 2 + 2 M 1 t 0 ( r + s ) + 4 � 3 M 3 3 t 3 0 − (2 M 1 + r + s ) t 2 f 2 ( r, s, M 1 ) = 0 f 2  � 1 r ( M 1 − t 0 ) 2 − 1 3( M 1 − t 0 ) 3 f 3 ( r, s, M 1 ) = f 2 1  r ( M 1 − t 0 ) 2 + 1 3( M 1 − t 0 − r ) 3 − 1 � 3( M 1 − t 0 ) 3 f 4 ( r, s, M 1 ) = f 2  � 1 s ( M 1 − t 0 ) 2 − 1 3( M 1 − t 0 ) 3 f 5 ( r, s, M 1 ) = f 2 1  s ( M 1 − t 0 ) 2 + 1 3( M 1 − t 0 − s ) 3 − 1 � 3( M 1 − t 0 ) 3 f 6 ( r, s, M 1 ) = . f 2

Expected Transmission Time Phone-to-AP Taking a similar approach (see AP’s as nodes stuck in hotspot mode) 1 1 M 2 r M 2 2 − ( M 2 − t 0 ) 2 ⇤ 2 f ( M 2 − t 0 ) 2 ⇥ = + I ( M 2 ≥ 0) + I ( t 0 ≤ M 2 <f − r ) T 2 2 f f M 2 1 2 ( M 2 − t 0 ) 2 − ( M 2 − f + r ) 2 ⇤ ⇥ + I ( M − 2 ≥ f − r ) + I ( M 2 <t 0 ) 2 f 2 f

Expected Transmission Time Phone-to-AP Taking a similar approach (see AP’s as nodes stuck in hotspot mode) 1 1 M 2 r M 2 2 − ( M 2 − t 0 ) 2 ⇤ 2 f ( M 2 − t 0 ) 2 ⇥ = + I ( M 2 ≥ 0) + I ( t 0 ≤ M 2 <f − r ) T 2 2 f f M 2 1 2 ( M 2 − t 0 ) 2 − ( M 2 − f + r ) 2 ⇤ ⇥ + I ( M − 2 ≥ f − r ) + I ( M 2 <t 0 ) 2 f 2 f Finally, given , the optimal mode toggling T 1 , T 2 2 , β , γ schedule is solved for using off-the-shelf solver. h r ∗ , s ∗ i

Implementation On Android Galaxy Nexus and Nexus S phones Using Java Reflection, no rooting is required Driving data: GPS trajectories & car OBD-II readings Measured mode switching overhead and communication range, and tested functional system in practice

Large-Scale Simulation

Large-Scale Simulation MSRA T-Drive taxicab dataset Central Beijing (50 km x 50 km) Feb 2~8, 2008 9211 taxicabs Assumed 10% WiFi coverage

Large-Scale Simulation MSRA T-Drive taxicab dataset Central Beijing (50 km x 50 km) Feb 2~8, 2008 9211 taxicabs Assumed 10% WiFi coverage Scenario G.N (Galaxy Nexus) and N.S (Nexus S) phones Cars collect driving data, share with each other Cars offload data to backend server when encountering APs

Large-Scale Simulation MSRA T-Drive taxicab dataset Central Beijing (50 km x 50 km) Feb 2~8, 2008 9211 taxicabs Assumed 10% WiFi coverage Scenario G.N (Galaxy Nexus) and N.S (Nexus S) phones Cars collect driving data, share with each other Cars offload data to backend server when encountering APs Schemes Baseline: no phone-to-phone communication Adaptive: system parameters are updated every hour using historical data Static: system parameters are computed using the first hour of data only

Simulation Results How does the mode switching overhead a ff ect optimal system e ffi ciency?

Simulation Results How does time-of-day a ff ect the optimal system e ffi ciency?

Simulation Results Improvement on transmission delay (mean and median) w/ phone-to-phone communication enabled v.s. w/o

Conclusion Our system enables vehicle-vehicle communications using off-the-shelf smartphones No change to existing infrastructure Transparent to end users Analytical formulation and results for optimal system efficiency Experiments show Over 80% system efficiency Significantly reduces data transfer delay time

Thanks