In this chapter we further investigated the transmission coordination problem in opportunistic routing and studied how the strict coordination requirement in ExOR can be greatly eased using network coding. First we reviewed MORE, a well-known state-of-the-art MAC-independent opportunistic routing protocol. After identifying the inadequacy of MORE in dealing with multicast and broadcast, we demonstrated how network coding can be integrated with opportunistic listening in wireless broadcast. In particular, we looked into a class of challenging problem—mobile content distribution (MCD) in VANET, where large files are broadcast proactively from a few APs to vehicles inside an interested area. To combat the lossy wireless transmissions in VANETs we leverage symbol level network coding (SLNC), which exploits symbol-level diversity to achieve better error-tolerance compared with traditional packet level network coding (PLNC). We then qualitatively characterize the advantages of SLNC compared with PLNC from two aspects, namely higher throughput and spacial-reusability. Using two typical MCD applications as examples—popular content distribution and live multimedia streaming, we present two novel push-based broadcast schemes—CodeOn and CodePlay, respectively. The common ideas underlying the two schemes include a prioritized relay selection algorithm that opportunistically maximizes the usefulness of transmitted content, and a simple transmission coordination mechanism that exploits the higher spacial reusability brought by SLNC. Compared with state-of-the-art pull-based contend distribution protocols, CodeOn and CodePlay both achieve significant performance gains. Through simulation study the gain can be partly attributed to the use of SLNC, and partly attributed to the new push-based protocol design. Finally, thanks to the use of opportunistic listening and network coding (especially SLNC), the challenging problem of designing a MCD protocol in VANETs is solved elegantly.
1 Unless otherwise stated network coding refers to packet-level network coding in the rest of the chapter.
2 This happens with high probability when the size of 
is large.
3 This assumption is valid in VANETs. The channel coherence time: 
, where 
is the doppler spread. With average relative speed v = 30 m/s, central frequency f0 = 5.9 GHz, Tc = 0.72 ms. Using the data rate 12 mbps in IEEE 802.11p, the time to send a 1 KB packet is Ttx = 0.68 ms. As Tc ≈ Ttx, consecutive received packets can be regarded as independent, so are the symbols in the same positions.
4 There exist error correction coding (ECC) techniques to enhance the error resiliency of packet transmission but they do not change the nature of the following derivation as they are still limited in error-correcting capabilities. On the other hand, ECC can also be added to SLNC (Katti et al. 2008).
5 Note that, in reality, the symbol errors may be correlated, which is related to the channel coherence time Tc. Then the actual difference between Ppe and Pse is smaller. Therefore, the gain we derived can be regarded as an upper bound.
6 We assume no node can receive more than one symbol or packet from different transmitters at the same time.
7 The property was original proved under random linear packet level NC, assuming 
is large. The same applies to SLNC, which is also based on random linear coding.
8 With packet-level broadcast, carrier sense is shown to work well under a two-transmitter setting in Brodsky and Morris (2009). Here we focus on a multi-transmitter setting instead, using SLNC.
9 We note that a similar segmentation approach has been used for solving different problems in previous works (Johnson et al. 2006; Li et al. 2009).