On Performance Analysis And Routing Technology For Ad Hoc Networks

The Ad Hoc network can be defined as a self-organizing wireless network which consists of nodes communicating with each other without any infrastructure or centralized administration. Since it can be rapidly deployed and is highly robust, the Ad Hoc network becomes an increas-ingly promising option for many critical applications, such as military troop communication, disaster relief, increasing the coverage of cellular networks, etc., and thus it is considered as an indispensable component among the next generation network architectures.Performance evaluation for Ad Hoc networks is of great concern, since it provides an important guideline for the practical network design and optimization, further promotes the development, application and commercialization of Ad Hoc networks. Routing protocol design is one of the most critical techniques which directly determine the data transmission. A well-designed routing protocol can improve the overall performance of an Ad Hoc network. The relevant work is supported by by the National S&T Major Project (No.2012ZX03004002-003),863 project (No.2007AA01Z217), National Science Foundation of China (No.61172079 and No.60972048). This dissentation first provides theoretical analysis for the evaluation on main performance metrics of Ad Hoc networks, such as delay, throughput and overhead, then with the help of our theoretical results, we propose several Ad Hoc network routing protocols. The main contributions of this paper are summarized as follows:The first part analyzes the delay performance of hop-limited (L≤2) Ad Hoc networks. Un-der such network scenarios, source nodes send multiple packet copies to different relay nodes, when one of these relay nodes moves to the location which is near to destination, it sends the copy to the destination. First, by modeling the packet delivery as a discrete stochastic process and computing the mean transition steps between its states, the upper bound of packet delivery delay is obtained, and the relationship between number of copies f and number of nodes N which optimizes the delivery delay is revealed asf= (?)(1/2N). Further, based on the mean residual service time, the packet queuing delay in its source node is theoretically derived and the packet end-to-end delay performance is obtained. Finally, the numerical results are provided to show the relationship of delay performance with the network size and the traffic load.The second part analyzes the capacity of buffer-limited Ad Hoc networks. When the buffer size of each node is fixed by K packets, with the fully consideration of the wireless interference and channel access competition, the node buffer is modeled as a G/G/1/K queue by employing the Queuing Theory. Based on this queuing model, a theoretical framework is developed to analyze the necessity conditions of the throughput upper bound and under which the packet mean arrival and service rates, and then obtain the per node’s throughput scaling law C as C = O(1 -1/K). Further more, Scheduling schemes which can achieve the throughput capacity are provided for both symmetric topology networks and asymmetric topology networks.The third part proposes a new concept of packet loss overhead for the buffer-limited Ad Hoc networks. i.e., when packets are lost in some relay nodes on a route due to their buffer overflow, the bandwidth used for the previous transmission of these discarded packets is wasted and thus defined as the packet loss overhead. By analyzing the conditions of optimal route in buffer-limited Ad Hoc networks, a theoretical framework is developed to evaluate the packet loss overhead for any given routing protocols. The simulation results show the existing routing protocols still have a great potential to reduce the overhead and improve the network throughput.The forth part proposes several efficient routing protocols for Ad Hoc networks are proposed. First, a traffic-aware routing protocol TACR is proposed. TACR utilizes the Wiener-based pre-diction methods to perceive the evolution of traffic distribution on the network, and then adjust the current routes timely. Meanwhile, a distributed Q-learning mechanism is introduced to dy-namically adjust the route lifetime, which improves the routing adaptability for different traffic distributions. Then, an end-to-end congestion control routing protocol termed as ECCR is pro-posed. By sensing the hops an congestion information of available paths, ECCR improves the network performance. Based on ECCR and combining the advantages of single-path and multi-path routing strategies, a mixed-path routing protocol termed as M-ECCR is further developed. Finally, based on the TACR and ECCR, a new load-balancing cognitive routing protocol NL-BCR is proposed. NLBCR utilizes the Wiener prediction to perceive the evolution of traffic distribution, and by adjusting the local current routes pro-actively, it releases network conges-tion without increasing the route length excessively. Meanwhile, NLBCR reconfirms the current route lifetime to avoid the unnecessary route rebuilding, thus the control overhead can be re-duced largely. The simulation results indicate that NLBCR can reduce routing overhead to a considerable extent, as well as improve the network throughput and decrease the end-to-end delay.