Kevin C. Lee, Uichin Lee, Mario GerlaKevin C. Lee, Uichin Lee, Mario GerlaNetwork Research Lab, UCLANetwork Research Lab, UCLA

{kclee,uclee,gerla}@cs.ucla.edu{kclee,uclee,gerla}@cs.ucla.edu

TO-GO: TOpology-assist Geo-OpportunisticTO-GO: TOpology-assist Geo-OpportunisticRouting in Urban Vehicular Grids Routing in Urban Vehicular Grids

TO-GO: TOpology-assist Geo-OpportunisticTO-GO: TOpology-assist Geo-OpportunisticRouting in Urban Vehicular Grids Routing in Urban Vehicular Grids

Motivation: unreliable wireless channel nature in vehicular urban gridsMotivation: unreliable wireless channel nature in vehicular urban gridsMotivation: unreliable wireless channel nature in vehicular urban gridsMotivation: unreliable wireless channel nature in vehicular urban grids

Geographic Routing in Vehicular Urban GridsGeographic Routing in Vehicular Urban GridsGeographic Routing in Vehicular Urban GridsGeographic Routing in Vehicular Urban Grids

Nodal planar graph: Recovery mode is expensive

Exploit spatial diversity for robust networking

Unreliable delivery, packet loss: Conventional geographic routing lack required robustness

Greedy Perimeter Coordinator Routing (GPCR)Greedy Perimeter Coordinator Routing (GPCR)Greedy Perimeter Coordinator Routing (GPCR)Greedy Perimeter Coordinator Routing (GPCR)

SRC

DST

• Greedy forwarding: forward a packet to its neighbor closest to the destination• But greedy forwarding is unreliable due to channel errors

Obstacles

Mobility

SRC

DST

DST

Road topology: use greedy forwarding in a road segment

SRC

GpsrJ+: 2-hop Junction PredictionGpsrJ+: 2-hop Junction PredictionGpsrJ+: 2-hop Junction PredictionGpsrJ+: 2-hop Junction Prediction

SRC

DST

GPCR: always forwards packets to a junction node

SRC

DST

2-hop neighbor info based prediction enables a forwarder to decide whether to bypass junction area

GpsrJ+: bypass junctions if there is no need to switch directions

Bypass junction

Junction nodes

Simulation Results Simulation Results Simulation Results Simulation Results

Exploit spatial diversity in urban gridsExploit spatial diversity in urban gridsExploit spatial diversity in urban gridsExploit spatial diversity in urban grids Forwarding set selection methodForwarding set selection methodForwarding set selection methodForwarding set selection method

TO-GO: Geo-Opportunistic Routing in Vehicular Urban GridsTO-GO: Geo-Opportunistic Routing in Vehicular Urban GridsTO-GO: Geo-Opportunistic Routing in Vehicular Urban GridsTO-GO: Geo-Opportunistic Routing in Vehicular Urban Grids

Existing schemes only considered the direction to the destination:

less efficient in vehicular urban grids

SRC

DST

SRC

DST

TO-GO: forwarding direction is determined using GpsrJ+ prediction

Finding a max set w/o hidden terminals:

clique problem NP-hard

• Choose a set that can hear the target• Find a neighbor node with largest neighbors• Filter its neighbors to avoid hidden terminals

Largest

Target

Wireless-free channelWireless-free channelWireless-free channelWireless-free channel

• To-Go obtains consistently high PDR with lower hop count• At node 150, high density helps GPSR and GPCR in delivery• GPCR always stopping at junctions hurts its delivery• GpsrJ+ has highest PDR b/c it is oblivious to hidden terminals

Wireless error-prone channelWireless error-prone channelWireless error-prone channelWireless error-prone channel

• Higher standard deviation, higher channel errors• TO-GO’s PDR remains high while GpsrJ+ decreases• GpsrJ+ hop count: [5.05, 7.47]; TO-GO hop count: [5.85, 8.35]