Techniques to mitigate traffic overload and protocol inefficiencies in mobile ad hoc networks

A mobile ad hoc network (MANET) is a collection of wireless devices moving in seemingly random directions and communicating with one another without the aid of an established infrastructure. Multi-hop communication between two distant nodes is achieved by other similar nodes acting as intermediate routers. Since there is no inherent traffic admission control, MANETs are likely to operate in a congested state, due to high traffic overload.; Owing to shared, half-duplex wireless channels and frequent node mobility, establishing and maintaining routes among communicating nodes is a challenge. Several routing protocols have been proposed specifically for ad hoc networks to handle frequent route breaks. However, extensive studies in literature and our own work indicate that the ad hoc routing protocols designed to excel, for example, in dense, low-mobility networks do not work well in sparse, high-mobility networks. Also, routing protocols may perform well at low to medium loads, prior to network saturation, but they do not sustain their performance for high loads.; This dissertation addresses these problems and provides several general techniques that can be used to augment existing routing protocols by changing the logic or algorithm used. We also provide general solutions that can correct an existing protocol's deficiencies without modifying the existing protocol. These solutions are implemented at the medium access control (MAC) sub layer and can be beneficial to many on-demand routing protocols.; We find that the IEEE 802.11 MAC protocol, as defined, is overcautious and prevents even lightly exposed nodes from communicating with their neighbors. Our modification to mitigate this weakness improves the throughputs of commonly used on-demand routing protocols by 10--20%.; As the network load increases, the packet delays increase, and routes are broken frequently due to exposed nodes. The former increases the route repair time, and the latter increases the control overhead, especially for on-demand routing protocols that use network-wide flooding to repair routes. We illustrate this for the most commonly used routing protocol, AODV, which loses throughput beyond saturation.; To mitigate the rapid loss of throughput by AODV, we propose two solutions: (a) removing duplicate control packets waiting at the MAC layer for transmission, and (b) a dynamically adaptive route repair timer to estimate the time needed to repair a route. The latter requires protocol changes, while the former does not. Using simulations, we show that both techniques enable AODV to perform gracefully under traffic overload conditions.; Under traffic overload, the number of false route breaks---due to temporarily non-response intermediate node in routing paths---is high. While the next hop is in the communication range, route breaks caused by transmission failures are called false route breaks. These false route breaks increase the routing overhead and cause performance degradation after the point of saturation is reached. Similarly, some routing protocols are prone for stale routes. Data packets that use stale routes consume network bandwidth for a few hops before they reach a dead-end and they are dropped, creating real route breaks. To reduce the false and real route breaks, we propose next hop status prediction techniques. Using prediction prior to transmission (pre-prediction) can reduce the effects of real route breaks. When the prediction is used after a transmission failure, the number of false route breaks may be reduced. Using two on-demand routing protocols, the AODV and the DSR, we show that our prediction schemes are effective, especially in high-traffic loads.