Routing over heterogeneous radio networks in mobile ad hoc networks

Abstract

There will be a need to configure nodes with multiple radios, as ad hoc networks are being deployed. The applications tend to have high bandwidth requirements. At the same time, the deployment scenario may result in range requirements that can not be met by Super High Frequency (SHF) or Very High Frequency (VHF) radios. An alternative is then to use multiple radios, where the radios differentiate in transmission range, bandwidth or by other radio characteristics. The aim of this dissertation is therefore to study various techniques to utilize heterogeneous radio networks, and for improved packet delivery within various configurations of mobile ad-hoc networks.

Wireless mobile ad hoc networks are commonly referred to as autonomous networks where the involved nodes organize themselves. The operations of ad hoc networks are neither requiring a preplanned infrastructure nor wired base-stations. A node is free to move, join or leave the network randomly. The nodes use their own wireless interfaces to both originate traffic, receive traffic and to convey traffic. Hence an ad hoc network consists of nodes collaborating in establishing a network enabling communication between the nodes.

Routing in heterogeneous radio networks was both the starting point, and the last work carried out in this dissertation. The work started by analyzing the Ad-hoc On Demand distance Vector (AODV) in a dual radio network environment. One radio was configured with high bandwidth, but a short transmission range, whereas the other was configured with low bandwidth and a long transmission range. The aim of this work was to investigate whether short-range but high bandwidth links would increase the overall capacity of the network. High bandwidth links were first studied by delaying route requests over long transmission range and low bandwidth links. Additional delay was also used to balance the traffic load. By delaying the route requests over the long range links, an improved utilization of the radio resources was demonstrated. The results show an increased packet delivery ratio compared to the commonly used shortest path routing methodology.

The proactive routing protocol Optimized Link State Routing protocol (OLSR) was also analyzed in a heterogeneous radio network setting. It was demonstrated that the overall traffic throughput increased when the routing took into consideration the observed traffic load. It was concluded that long transmission range and low bandwidth links were preferable at low traffic load, while high bandwidth links were preferable at high traffic load. At low traffic load, the low bandwidth radios provide sufficient bandwidth and consistent routes. However, the low bandwidth radios did not provide sufficient capacity at high traffic loads. The cost of the increased utilization of high capacity radios was more inconsistent route entries. Hence, low capacity radios were preferably used at low traffic load, while high capacity radios should be used to reduce the load over the low capacity radios.

The work also addressed various techniques for improved packet delivery under different network configurations and traffic load. Due to node mobility, stale route entries cause packet loss. The performed study showed that by changing the traditional queuing method, and by providing packets a second chance, an improved packet delivery was achieved. The change in queue methodology consisted of assigning the next-hop address after queuing instead of before queuing. A second chance was also provided to the otherwise lost packets by alternative routing. As a result, the otherwise lost packets due to mobility were preserved.

The alternative queue method was also carried out in an experimental test-bed using Linux software routers. The simulation and test-bed results showed similar trends, but deviated in terms of the number of lost packets. The test-bed did also disclose some common protocol miss configurations. Broadcast is commonly used for the dissemination of control packets. In order to receive the broadcast packets, the interfaces are traditionally configured with a netmask wide enough to receive these packets. This miss configuration resulted in data packets being tried transmitted, though no 32 bits route entry was available. Upon a missed 32 bits route entry, the packets were forwarded due to the longest prefix match, which was the interface netmask. Hence, data packets were not experiencing route error when broadcast was used for the dissemination of the control packets.