Connectionless approach -- a localized scheme to mobile ad hoc networks
Kien A. Hua, Yao-Hua Ho · 2009
One important problem with Mobile Ad hoc NETwork (MANET) is the routing protocol that needs to work well not just with a small network, but also sustain efficiency and scalability as the network gets expanded and the application transmits data in greater volume. In such an environment, mobility, channel error, and congestion are the main causes for packet loss. Due to mobility of mobile hosts, addressing frequent and unpredictable topology changes is fundamental to MANET research. Two general approaches have been considered: connection-oriented approach and connectionless-oriented approach. In the former, the emphasis is on how to reconnect quickly with low overhead when a broken link occurs. Examples of this approach includes [5], [9], [10], [16], [26], [28], [29], [34], [44], and [45]. In contrast, connectionless-oriented approach focuses on minimizing the occurrence of broken links. We proposed one such scheme called Connectionless Approach (CLA) [21] and [22]. In CLA, the network area is divided into non-overlapping grid cells, each serving as a virtual router. Any physical router (i.e., mobile host), currently inside a virtual router, can help forward the data packet to the next virtual router along the virtual link. This process is repeated until the packet reaches its final destination. Since a virtual link is based on virtual routers which do not move, it is much more robust than physical links used in the connection-oriented techniques. Simulation results in our previous works [21] and [22], based on GloMoSim [60], indicate that CLA performs significantly better than connection-oriented techniques (i.e., AODV, DSR, LAR, GRID, TMNR, and GPSR). The contribution of this work consists of investigating and developing new Connectionless-Oriented Approach for Mobile Ad Hoc Network. Two of the greatest impacts of this research are as follows. First, the new approach is targeted towards robustly support high mobility and large scale environment which has been adapted for vehicle-to-vehicle environment in [20]. Second, the detailed simulations which compare eight representative routing protocols, namely AODV, DSR, LAR, GRID, TMNR, GPSR, CBF, and CLA, under high-mobility environments. As many important emergent applications of the technology involved high-mobility nodes, very little is known about the existing routing methods perform relative to each other in high-mobility environments. The simulation results provide insight into ad hoc routing protocols and offer guidelines for mobile ad hoc network applications. Next, we enhanced and extend the connectionless-oriented approach. The current connectionless-oriented approach, however, may suffer from packet drops since traffic congestion is not considered in the packet forwarding policy. We address this weakness by considering the connectionless-oriented approach with a collision avoidance routing technique. After that, we investigate techniques to enforce collaboration among mobile devices in supporting the virtual router functionality. Many works have been published to combat such problem - misbehaving nodes are detected and a routing algorithm is employed to avoid and penalize misbehaving nodes. These techniques, however, cannot be applied to the connectionless-oriented approach since any node in the general direction towards the destination node can potentially help forward the data packets. To address the security and cooperation issues for connectionless-oriented approach, we introduce a cooperation enforcement technique called 3CE (3-Counter Enforcement). In addition, wireless mesh networks have become increasingly popular in recent years. Wireless mash network (WMNs) are collection of mesh clients and mesh nodes (routers), with mesh nodes forming the backbone of the network and providing connection to the Internet and other network. We propose a paradigm that combines virtual routers and mesh nodes to create a hybrid network call VR-Mesh Network. This hybrid network can reduce number of mesh node needed without decrease the performance of the network. (Abstract shortened by UMI.)