Real-time video transmission over ad hoc networks faces many challenges including low bandwidth, long end-to-end
delay, high packet loss rate, frequently changing topology and limited-powered mobile nodes. This paper presents an
effective real-time video transmission scheme and improves implementation of DSR (Dynamic Source Routing)
protocol. We set up a test-bed by using DSR routing in the IP layer, and an application transmitting video stream over
UDP protocol. We get a continuous JPEG image stream from a ZC0301p web camera and split each image into small
blocks according to the MCU (Minimum Coding Unit) borderline. The strong point of splitting JPEG image is that IP
layer fragmentation can be avoided so we can determine which part of data in the frame gets lost to do loss recovery at
the receiver. By using JPEG image stream, the video encoding complexity is reduced, which can save computing power
of mobile nodes compared with MPEG and other Multiple Description Coding (MDC) methods. We also improve
implementation of DSR to make it suitable to transfer real-time multimedia data. First different priorities are given to
different traffic classes in DSR routing. Second the route maintenance scheme is modified to decrease overhead and link
failure misjudgments. We carry out two experiments both indoors and outdoors using six mobile nodes. The first is to
transmit continuous JPEG images using our former DSR implementation according to DSR draft. The second is that we
split JPEG images into blocks and then transmit them using improved DSR implementation. Results show the latter
gives better video stream fluency and higher image quality.
In this paper, Extreme Value Theory (EVT) is presented to analyze wireless network traffic. The role of EVT is to allow the development of procedures that are scientifically and statistically rational to estimate the extreme behavior of random processes. There are two primary methods for studying extremes: the Block Maximum (BM) method and the Points Over Threshold (POT) method. By taking limited traffic data that is greater than the threshold value, our experiment and analysis show the wireless network traffic model obtained with the EVT fits well with that of empirical distribution of traffic, thus illustrating that EVT has a good application foreground in the analysis of wireless network traffic.
Significant TCP unfairness in Ad Hoc wireless networks has been reported during the past several years.
proposed a network layer solution called Neighborhood Random Early Detection (NRED) scheme to enhance TCP
fairness in Ad Hoc wireless networks. In NRED, the concept of neighborhood is introduced. So the RED mechanism is
extended to the distributed neighborhood queue, which is the aggregation of local queue in one's neighborhood. NRED
adopt a passive measurement technique to detect the early congestion of a neighborhood. However, NRED by measuring
channel utilization rate is an over-layer solution and hardly to implement in practice. As it is known, packet delay
increases when the wireless channel is very busy and the overall traffic load exceeds the capacity of the channel. Thus
the packet delay can reflect whether or not the channel is busy. For each packet's transmission, the more delay, the more
severe congestion and competition. We believed that the delay of data could reflect the congestion of shared link
promptly. This paper proposes a scheme based on MAC delay to detect congestion and to notify the nodes which use too
much channel dropping their packets and give the expressed node chance to transmit. We analyze the average packet
delay on IEEE 802.11 DCF which is represented by a Markov model. Based on the relationship between the MAC delay
and number of competitors, whether there exist severe competition can be found.
With the fast development of ad hoc networks, SIP has attracted more and more attention in multimedia service. This paper proposes a new architecture to provide SIP service for ad hoc users, although there is no centralized SIP server deployed. In this solution, we provide the SIP service by the introduction of two nodes: Designated SIP Server (DS) and its Backup Server (BDS). The nodes of ad hoc network designate DS and BDS when they join the session nodes set and when some pre-defined events occur. A new sip message type called REGISTRAR is presented so nodes can send others REGISTRAR message to declare they want to be DS. According to the IP information taken in the message, an algorithm works like the election of DR and BDR in OSPF protocol is used to vote DS and BDS SIP servers. Naturally, the DS will be replaced by BDS when the DS is down for predicable or unpredictable reasons. To facilitate this, the DS should register to the BDS and transfer a backup of the SIP users' database. Considering the possibility DS or BDS may abruptly go down, a special policy is given. When there is no DS and BDS, a new election procedure is triggered just like the startup phase. The paper also describes how SIP works normally in the decentralized model as well as the evaluation of its performance. All sessions based on SIP in ad hoc such as DS voting have been tested in the real experiments within a 500m*500m square area where about 30 random nodes are placed.
This paper investigates saturation and nonsaturation load throughputs of node based on IEEE 802.11 wireless ad hoc network protocol in presence of selfish node. To analyze the throughput of nodes, an extended two-dimension Markov model was used and a general analytical solution was derived for DCF that may be used to find throughput under various traffic loads. Meanwhile, we have done extensive simulation using Qualnet to validate our analytic results. The analytic and simulation results matched well, which revealed three interesting insights: 1) The selfish node can maximize its throughput by adopting selfish behavior. And with the increase of selfish node, the throughput of selfish node decreases. 2) With the increase of initialized contention window size of selfish node, the throughput obtained by selfish node decreases. 3) The effect taken by the selfish behavior increases with the increase of the traffic load of node. When the traffic load is saturation, the throughput that the well-behaved node can get nearly approaches to zero, which is a very undesirable results.