| Abstract: |
Micro Aerial Vehicles (MAVs) are small flying robots often equipped with on-board cameras, GPS, and other sensors, and enable a manifold of novel civilian services. Surveillance of large
farmlands, 3D mapping of an area, as well as search and rescue (SAR) missions are just a few of the growing number of applications. To make use of the information gathered by MAVs, connectivity
between the MAVs and the ground stations has to be provided. The MAVs themselves may therefore create an aerial vehicle network. Ad-hoc networking solutions are available, yet, in MAV networks,
standard wireless communication technologies are challenged by frequent MAV movement, signal obstruction by the MAV frame, and antenna type as well as orientation influencing signal propagation,
that result intermittent connectivity. Here, traditional mobile ad-hoc routing protocols do not perform well as they require end-to-end connectivity, and pure delay-tolerant networking (DTN)
approaches are not optimized for MAV networks. At the same time, MAVs also provide unique opportunities ranging from availability of location and motion sensors to controlled mobility, that are
not available in conventional networks, and can benefit wireless communications. Overall, the dimensions of these challenges and opportunities are not a-priori clear. In this dissertation, we
propose MAV networking solutions following an empirical approach. We develop and make use of a testbed optimized for the SWARMIX research project on technology-enhanced search and rescue
operations. First, we investigate the single aerial link using a platform consisting of two popular types of MAVs, micro airplanes and quadcopters. We characterize the throughput of a link as a
function of distance between MAVs communicating through IEEE 802.11n, and further study the impact of major parameters on aerial communications. We also counteract signal obstruction by the MAV
frame and antenna-caused impairments by adapting the MAV platforms. Second, aerial link transmissions are optimized by introducing ferries and a location-aware transmission scheduling strategy
based on delayed gratification. The control mobility feature of MAVs is leveraged to efficiently transmit bulk data such as images on the single link under high mobility. Using the model, the
rendezvous point between sender and receiver is determined where the communication delay is expected to be minimal, and the sender MAV moves to this point and transmits the data. In practice, we
show that this strategy can substantially reduce delay and power consumption of MAVs. Third, we present a location-aware packet forwarding algorithm that combines end-to-end routing and delay
tolerant forwarding in a multi-hop aerial network. Due to non-random mobility of MAVs, two novel light-weight heuristics anticipating future locations are proposed that bring motion-awareness to
the algorithm. We support our analysis by a measurement campaign with a testbed of quadcopters and with a larger MAV fleet in simulation. Our experimental results reveal the practicality of the
algorithm and show that an anticipatory approach can alleviate inefficient forwarding and improve delay. Finally, we utilize the knowledge of mission available in MAV networks to propose a
mission-aware packet forwarding algorithm that - in addition to current location information - looks into the MAVs future waypoints to further optimize DTN packet forwarding. Evaluated in
simulation and in field tests, the algorithm shows promising results, in particular, in reducing the number of transmissions. |
