7.2 Mobile Content Distribution in VANETs
Vehicular communication has been a topic of great interest in recent years. Typically, a vehicular network consists of roadside units (RSUs), which are access points along the road, and on-board units (OBUs) mounted on vehicles, which can either communicate with the RSUs or with other OBUs in an ad hoc manner. Since the advent of dedicated short-range communications (DSRC) (Jiang et al. 2006); see also www.standards.its.dot.gov/Documents/advisories/dsrcadvisory.htm., IEEE 802.11p and IEEE 1609 standards (WAVE 2006), people have envisioned and designed numerous tempting applications of vehicular networks, ranging from safety warnings (Li et al. 2009) and intelligent navigation, to mobile infotainment (Lee et al. 2006). Among them, content distribution, especially “popular” multimedia content distribution to vehicles inside a geographical area of interest, is particularly attractive. Examples of such mobile content distribution (MCD) include live video broadcast of road traffic and conditions to vehicles driving towards it for intelligent navigation, which is especially useful during inclement weather; periodical broadcasts of multimedia advertisements for local businesses in a city to vehicles driving through a segment of suburban highway (like a digital billboard) and the dissemination of an accurate update of the GPS map about a city or a scenic area.
The challenges of providing a MCD service in VANETs are threefold. On the one hand, content distribution, especially multimedia content consisting of audio and video, requires high distribution rate and short delay. For live multimedia streaming, further quality of service requirement, such as the deadline for receiving a packet, which translates to smooth playback or stable downloading rate, is also important. On the other hand, wireless is the well-known shared and lossy medium with very limited bandwidth and throughput drops dramatically after a few hops in multihop wireless networks (Jain et al. 2005). Moreover, the high mobility of VANETs, leading to fast and unpredictable topological changes will further exacerbate the frequent packet losses and collisions.
Network coding is a common technique adopted in content distribution as an effective approach to improving the bandwidth efficiency and simplifying the protocol design. However, traditional network coding combines information at the packet level. Generally speaking, if a packet is received in error, then it is discarded by an intermediate node, which involves more transmissions and wastes the channel bandwidth. Under adverse channel environments such as the mobile VANET, the link-loss problem is worse than in static networks like wireless mesh networks (Torrent-Moreno et al. 2004; 2005; 2006). Because of the low packet reception success rates, using packet-level network coding (PLNC) is no longer sufficient to meet the requirements of some MCD services (e.g, multimedia streaming) that require high and stable data rate and short delay in VANETs (Park et al. 2006; Yang et al. 2010). In addition, protocols adopting PLNC also face the well-known hidden-terminal problem in MWNs, which is notoriously difficult to solve in multicast/broadcast (Dutta et al. 2009; Ni et al. 1999).
On the other hand, the benefit of network coding often tends to be offset by severe packet collisions due to lack of proper transmission coordination mechanisms among vehicles (Li et al. 2011). However, too much coordination again incurs problems as we have mentioned before: the overhead required by strict coordination per se would in turn negatively affect the protocol performance. Obviously, adopting a broadcast-tree-based method is impractical because it needs to collect the global topology information in real-time. There is therefore a choice about the level of node coordination that is necessary. We contend that it is always beneficial to reduce the effort spent in coordination to the minimum amount.
Novel techniques are needed to address all of the above challenges. The proposed schemes in this chapter are based on symbol-level network coding (SLNC) (Katti et al. 2008). In contrast to traditional packet level network coding, SLNC allows intermediate nodes to combine packets at symbol level, where a symbol is typically composed of several physical layer symbols of a modulation scheme. Symbol-level network coding allows a node to recover correctly received symbols from erroneous packets. As symbol error rate is smaller than the packet error rate, in addition to the benefits one can gain from PLNC, SLNC provides better error tolerance and thus increased successful packet reception rate. A further study of SLNC also shows that SLNC in fact enables higher spacial reusability by allowing concurrent transmissions within shorter distances (Li et al. 2011; Yang et al. 2010), which enables much simpler node coordination than using PLNC.
In addition, by fully exploiting SLNC and opportunistic listening, a new push-based protocol design concept is proposed in this chapter. Specifically, the content source actively “pushes” information to nearby vehicles, while a dynamic set of temporary relay nodes is chosen to help further broadcast this information to all the other vehicles in the VANET. The source's responsibility to guarantee reception and quality of service for the whole network is distributed to each relay node, which only needs to ensure the reception of its neighboring vehicles (Yang et al. 2009; 2011). The node coordination is kept to the minimum, since whoever receives the largest amount of useful contents within a local range will be chosen as a relay node, and the relay nodes simply compete to access the wireless channel randomly based on carrier sensing, as that in IEEE 802.11.
In this chapter, we will demonstrate the effectiveness of our proposed methods using two example MCD applications: popular content broadcast and live multimedia streaming. Specific schemes designed for each of these two applications are presented, where the concept of opportunistic listening are exploited in different ways.
7.2.1 Model and Assumptions
In this chapter, we consider the basic network architecture for the MCD service, which is illustrated in Figure 7.2. We assume there are one or more access points (APs, or road side units) deployed in the area of interest (AoI), which can be either a highway segment or an urban area. The content provider (e.g. a citywide traffic administration bureau) would like to distribute some content/multimedia content through the APs to vehicles inside the AoI. However, due to the deployment cost and limited communication range of APs, the entire AoI is not fully covered by APs. For vehicles outside the direct communication range of an AP, they form a VANET and cooperatively collect/distribute the content.
Figure 7.2 The architecture for MCD. Inside the AP coverage, AP broadcasts and vehicles receive; outside the AP coverage, vehicles distribute their received contents cooperatively. Reproduced by permission of © IEEE 2011.
Each vehicle is equipped with an on-board unit including a wireless transceiver (single radio). The wireless interface operates on multiple channels (www.standards.its.dot.gov/Documents/advisories/dsrcadvisory.htm; Jiang et al. 2006). To model the coexistence of safety and commercial applications, we consider two representative channels. The control channel is used to broadcast safety messages, which may contain vehicles' locations, speeds etc.; one service channel is dedicated for MCD. In order to guarantee the quality of service of safety messages (the interval between two consecutive safety messages should be smaller than 100 ms—et al. 2009), time is divided into periodical, 100 ms slots and all vehicles and APs are synchronized to switch simultaneously between the control channel and service channel. The utilization of time and channels is depicted in Figure 7.3. Although there are advanced MAC protocols that dynamically adjust the time shares of control channel and service channel for better service (Mak et al. 2009), we fix it to 1/2:1/2 for simplicity.
Figure 7.3 The time and channel utilization of each vehicle and each AP. Reproduced by permission of © IEEE 2011.
In the control channel, each AP and each vehicle broadcasts one beacon message in each slot. When a vehicle is in the range of an AP, it merely listens to the AP's content broadcast in the service channel; otherwise, it may share its received content with neighboring vehicles cooperatively. Vehicles outside the AoI do not involve in content distribution.
In addition, we assume all vehicles are equipped with Global Positioning System (GPS) devices, from which vehicles obtain their real-time locations and synchronize their clocks (error smaller than 100 ns). GPS devices have a low cost and are available to most drivers nowadays. When vehicles are temporarily out of satellite coverage, they can use auxiliary techniques to determine their location, and rely on their own hardware clocks. Note that GPS time synchronization is required by the IEEE 1609.4 standard for multichannel operations (WAVE 2006).