Maximizing throughput of relay networks using UAV's

This thesis investigates two effective system methods for maximizing throughput in practical networks where low-altitude low-speed Unmanned Aerial Vehicles (UAVs) serve as relays for ground nodes. I consider the task of using multiple UAVs to relay packets between two distant ground nodes, and I seek to maximize the throughput for delay-tolerant applications subject to the constraint of application allowable delay. I also focus on the use of IEEE 802.11 radio equipment. My first approach is to exploit controlled mobility. I propose a simple Load-Carry-and-Deliver (LCAD) network, discuss necessary optimality conditions for the network, evaluate its performance in field experiments, and derive an empirical performance model using the experimental results. The results show that the communication performance using IEEE 802.11 radio with aerial node can be predicted with good accuracy, and for maximizing throughput we can use the performance model to design flight path and schedule communication. My second approach is to use multiple radios in parallel. Interference problems are inevitable when multiple radios are transmitting simultaneously. I propose three methods of parallel channel use in a multi-hop network, discuss interference issues each is likely to experience, and evaluate their performance in field experiments. Among the three, the Time-Division Multiplexing (TDM) method achieves the best performance in the experiments because it incorporates several techniques designed to mitigate interference. These techniques can be employed in a relay network for effective use of multiple radios in parallel. For maximizing throughput in a relay network with multiple UAVs, I compare two relay paradigms, MHR ( Mutli-Hop Relay o) and LOAD. Under simplified assumptions, I analyze their delay-throughput trade-offs, and highlight the areas in which one paradigm can outperform the other. In general, MHR can achieve a higher throughput when the allowable delay is small and when only few UAVs can be used; LCAD, on the other hand, can achieve a higher throughput when the allowable delay is large and when more UAVs can be used. I also compare some system properties of these two paradigms and argue that some properties of LCAD make it a better paradigm for use in practical relay networks.