2.1. VADD
The high mobility of vehicles and the rapid-changed network topology result in the vehicular network disconnecting frequently. This occurs more easily when the vehicular density is sparsely (e.g., not rush hour).
VADD [6] targets at solving such problem by exploiting the opportunities of intermittent connection between moving vehicles. Different from connection-oriented routing protocols, such as AODV and DSR, VADD does not need the exchanges of routing control messages to establish a path with end-to-end connection, maintain the available paths and repair the broken path. Therefore, the routing cost can be significantly reduced. VADD utilizes the idea of carry-and-forward scheme which is well known in the area of Delay Tolerance Networks (DTN), to pass packets while the network connection is unserviceable. A sender vehicle can just carry the unforwardable packets until it meets other vehicles in transmission range, and relays the buffered packets to those neighboring vehicles. This is of great help to delay-tolerant applications.
By the traffic statistics information provided from the preloaded digital maps and by the coordinates obtained from the GPS device, a stochastic delay model is built accordingly. The delay model estimates the data-delivery delays to assist the road segment (or intersection) selection, and finds the minimum delivery delay path further.
2.2. VANET routings for video streaming
The authors of V3 [10] provided a vehicle-to-vehicle live video streaming architecture for highway scenarios. Vehicles in the destination region act as video
sources to capture and send the videos which contain the surrounding information to the requester vehicles. To accomplish this goal, a video source trigger sub-system and a video data transfer sub-system are proposed as the main functionalities of V3. The former uses a signaling mechanism to let the requester vehicles can possess the freshest video information by continuously triggering the video sources to send video back; and the latter applied the carry-and-forward scheme to efficiently transfer video packets by using some forwarder selection approaches. However, the authors did not use real video data in the simulation.
[16] also proposed two routing protocols, the sender-based forwarding (SBF) and receiver-based forwarding (RBF), to send video packets in highway environment.
SBF chooses the vehicle which is closest to the destination vehicle to be the packet forwarder, this approach is also known as greedy forwarding. In addition, RBF follows the concept of contention-based forwarding [17] to elect forwarders. The sender vehicles apply broadcast to transmit video packets instead of unicast. The neighboring vehicles which have received the packet will delay a short period before sending to the link layer according to the delay principle. If the node with shortest delay has broadcasted the packet out, then other nodes overhearing the transmission will discard the packet. According to their work, we can observe that the RBF approach has no need of any control packets and achieves better performance than SBF because of no extra overhead. Nevertheless, [18] indicated the delay time should be large enough to distinguish the best forwarding candidate and other candidates.
Furthermore, the delay time should be short to avoid unnecessary waiting time. How to design a delay function which is suitable for the traffic with bursty nature is not trivial.
For the urban environment, [19] proposed a cross-layer path selection algorithm to send video packets with the help of road side units (RSUs). The authors adopted a
stochastic mobility model which divides a road segment into the front, the middle and the end parts to consider, and queuing theory was used to calculate the connectivity of a road segment. The RSU plays a role of the video source; it sends packets to the destination vehicle via the relaying of moving vehicles. Vehicles periodically broadcast their location information and the RSU can accordingly plan a best routing path to deliver video packet by using video distortion as the concerned routing metric.
Different to this work, the study of this thesis does not rely on the help of RSUs, we try to follow the similar way of VADD to deliver video packets.
2.3. Distortion-based video streaming in MANET
To provide better video streaming services in mobile computing devices and wireless networks, researchers start to consider the relationship between the characteristics of the video and of the wireless network environment in their research.
Such researches aim at offering the end users better perceived video quality. Because the video quality perceived by each user is probably not the same, Peak Signal-to-Noise Ratio (PSNR), the most widely used metric to measure the quality of received video, can helpfully judge the quality of video:
(1)
and the Minimum Square Error (MSE) is:
(2)
The purpose of PSNR is to measure the difference between the original video frame and the processed video frame. The higher PSNR value, the better video quality, and vice versa.
Stuhlmuller, et al. [20] provided an empirical rate-distortion model for a hybrid motion compensated video encoder. The authors analysis the codec structure of H.26x
and MEPG-x series video coding standards to induce the general form of distortions are in the form of MSE and can be converted to PSNR values by
[13] proposed a distortion-based routing path selection algorithm for static ad hoc networks. The work applied the rate-distortion model as mentioned earlier, and
The routing algorithm proposed by [14] and [15] considers not only the channel and packet expiration issues but also the influences of the video encoding parameters.
They based upon the factors to find an optimal routing path for individual packets.
The distortion calculation approaches are basically similar to [13] (although the used distortion models are different).