1. When does opportunistic routing make sense? Rahul C. Shah, Jan Rabaey University of California, Berkeley Sven Wiethölter, Adam Wolisz Technical University, Berlin

2. Problems of geographic routing Fading channel Inter-packet times are longer than the channel coherence time Duty cycling of nodes Geographic Routing Advantages of geographic routing: Tolerant of node failures/movements Very small routing table sizes

3. Channel Variations Over Time Broadcast Success Rate (%)

4. Networking Protocol Stack Application Network (Routing, Addressing) Data Link (MAC, Sleep discipline, power control) Physical Next hop node information Observes network connectivity Observes neighbor connectivity Observes channel quality Enhance information being exchanged between the layers to maximize power efficiency

5. Exploiting Spatial Diversity: Opportunistic Routing Current node destination Nodes know: Their own location The destination location MAC chooses next hop based on connectivity Network layer specifies forwarding region

6. Exploiting Spatial Diversity: Opportunistic Routing Current node destination Nodes know: Their own location The destination location Network layer specifies forwarding region MAC chooses next hop based on connectivity

7. Exploiting Spatial Diversity: Opportunistic Routing Current node destination Nodes know: Their own location The destination location Network layer specifies forwarding region MAC chooses next hop based on connectivity

8. Exploiting Spatial Diversity: Opportunistic Routing Current node destination Nodes know: Their own location The destination location Network layer specifies forwarding region MAC chooses next hop based on connectivity

9. Opportunistic TICER for Medium Access Extension of TICER (Transmitter Initiated Cycled Receiver) Provides ability to choose among multiple candidate forwarding nodes Pseudo-asynchronous rendezvous scheme Pick the first node that responds to the RTS Assume channel is stationary over the RTS-CTS-Data-ACK exchange

10. Other Opportunistic Protocols Geographic Random Forwarding (M. Zorzi & R. Rao) –Uses geographic location of nodes to find best node –Divides forwarding region into priority regions –MAC protocol signaling is fairly complex Extremely opportunistic routing (S. Biswas & R. Morris) –Ranks forwarding nodes by number of hops –Sender specifies priority of receiving nodes in the packet MAC layer anycast (R. Choudhary & N. Vaidya) –Provides framework for choosing forwarding nodes at the MAC layer

11. Simulation Details Simulation was done in OMNeT++ Random placement of nodes in a square Poisson traffic was generated at the edges Circular radio range of 10m, interference range ~1.5 times radio range Compared with geographic routing Metrics of comparison: –Power consumption –End-to-end delay –Goodput Precision of mean within 5%, confidence level of 95%

12. Varying Node Density Average number of neighbors per node Power consumption per node (mW) Average end-to-end delay (sec)

13. Channel Quality Variations Power per node/Pkt delivery fraciton (mW) Pareto distribution shape parameter End-to-end delay (sec)

14. Varying Traffic Rates Power consumption per node (mW) Packet inter-generation time (sec)

15. Change in Wakeup Rate End-to-end delay (sec) Power per node (mW) Node wakeup rate

16. Optimum Wakeup Rates Avg. no. of neighbors Wakeup rate per node Duty cycle per node Avg. # of nodes awake in fwd. region 6.33.31.6%0.04 8.83.31.6%0.06 11.32.81.4%0.06 13.82.21.1%0.06 16.32.21.1%0.07 18.82.01.0%0.08 Less than one node needs to be awake within the forwarding region!

17. When does opportunistic routing make sense? (compared to geographic routing) Node densities > 9-10 neighbors/node Medium to bad channel quality Low traffic scenarios (<1 packet/second per node) Improvements –10-20% lower power consumption –40% lower delay suffered by the packets –Fewer retransmissions required  higher goodput (10-15% higher for max of 3 retransmits)