|
Banca |
Prof. Leandro Buss Becker, Dr. – DAS/UFSC (presidente);
Prof. Aldebaro Klautau, Dr. – PPGEE/UFPA;
Prof. Eduardo Souto, Dr. – Icomp/UFAM;
Prof. Carlos Barros Montez, Dr. – DAS/UFSC. |
| Abstract: Performing means for exchanging messages with stable connections in missions composedof multiple unmanned aerial vehicles (UAV) is a complex task. The variations in UAVdistances from each other, considering their trajectories, and the medium dynamic factorsimpose difficulties that must be properly addressed. In addition, the use of different types-of-service (ToS) such as voice, data, and video are increasingly present in the executionof applications involving networks composed of multiple UAVs. In this way, the reliabil-ity in the delivery of messages and link quality are important research challenges thatensure the exchange of different messages in a dynamic way, meeting the different ToSnetwork requirements to obtain traffic with quality of service (QoS). The use of hetero-geneous communication medium has shown gains in maintaining the connection amonghighly mobile nodes while increasing reliable transmission of data, as needed in MANETS,VANETs, and, more recently, in FANETs. In this context, this thesis proposes a heteroge-neous interface manager (IM) that is capable of improving communication in multi-UAVnetworks. Given a predefined set of available individual wireless interfaces, the proposedIM dynamically defines the best interface for sending messages based on on-flight condi-tions sensed and calculated dynamically from the wireless medium. The proposed IM issituated above the network link layer and contains a heuristic that decides in real-timebetween two or more wireless communication interfaces. It considers up-to-date monitor-ing data that represent the current state of the UAVs’ communication links within a givenenvironment. Currently, a decision-three (DT) with a sum of points heuristic is used asbasis to the proposed approach, and aims to support the decisions of IM. The heuris-tic accounts for parameters, such as the number of bytes received, number of bytes lost,throughput, received signal strength indication (RSSI), and signal to noise ratio (SNR).Firstly, the proposed IM is implemented using IEEE 802.11n and IEEE 802.11p wire-less interfaces employing different frequency bands to validate the decision tree heuristic.Secondly, the IM applies several single-band and multiband wireless local area communi-cation interfaces: IEEE 802.11n, IEEE 802.11p, IEEE 802.11ac, and IEEE 802.11ax, inorder to extend the IM decision possibilities. Lastly, a Naive Bayes classifier is attachedto the IM to perform a policy of weights according to network metrics. It computes thenetworks most critical points defining an adaptive solution according to the type of traffic.The IM is validated with simulations conducted using a realistic (and complex) simulationsetup based on the NS-3 network simulator, connected to the Gazebo or SUMO mobilitysimulators. This allows the UAV missions to be programmed in 2 D and 3 D mobility’sscenarios, and leave all communication aspects to be processed within NS-3. There wereconducted several experimental scenarios involving different numbers of UAVs, flying atdifferent speeds, traveling different distances and trajectories. The aim was analyzing theperformance of the communication interfaces applied homogeneously (with a single interface) and heterogeneously (using the proposed IM with different set of interfaces). The IMperformance was evaluated in terms of metrics from the medium-access-control (MAC)and physical layers, aiming to improve and maintain the connectivity between the UAVsduring the mission, and from the application layer, which targets the reliability in thedelivery of messages. Obtained results show that compared with the cases where a sin-gle interface is used, the proposed IM can increase the network throughput and presentsthe best proportion of transmitted and received packets, reception power (-60 dBm to-75 dBm), and loss (-80 dB to -85 dB), resulting in more efficient and stable networkconnections. Finally, this thesis also evaluated the performance considering voice, data,and video streaming ToS, highlighting the gains of the IM with and without ML classifier.The results showed that a combination of different interfaces in the IM decisions usingadaptive metric weights compose a powerful solution to maintain and increase the linkquality in U2U achieving message exchange over greater distances. |
)
