Deep-learning-based transport layer control for big data transmission over UAV swarming networks

Multi-Beam Smart Antennas (MBSAs) achieve concurrent communications in multiple beams, thereby providing higher throughput compared to regular directional antennas. Most of the transport layer control schemes used in MBSA-equipped networks are based on TCP, which are too conservative since they indiscriminately reduce the window size upon any kind of packet loss. To adapt to the features of the MBSA-based networks, this work proposes a balanced transport control strategy, which is based on the idea of Adaptive Batch Coding (ABC). The proposed ABC scheme resists random loss through redundant coding and copes with congestion loss via window shrink and retransmission. Using a cross-layer design, the coding scheme is adaptively adjusted according to the network conditions. A customized simulation system has been developed to comprehensively evaluate the performances of the proposed ABC protocol. Experimental results show that the proposed scheme overcomes the drawbacks of the traditional transport layer control methods and achieves higher good throughput and lower end-to-end delay in the MBSA-based networks. Furthermore, using deep learning networks, the coding parameters of the proposed ABC scheme are further optimized, thereby improving the communication performances. Meanwhile, to overcome the high computational complexity and large-scale broadcast of the current key management protocols, this work proposes a novel group key management protocol for multi-path wireless networks, which avoids high computation cost by employing a secret sharing algorithm based on Chinese Remainder Theory. The proposed scheme lets the information source choose the session key and piggybacks the key management messages onto routing communications. By this approach, all the nodes belonging to the session are authenticated and distributed the session key at the same time of routing. Therefore, it is more efficient and more suitable for the multi-path networks.

Grantor dc:publisher

University of Alabama Libraries

Year dc:date.issued

2019

Author dc:creator

• Mao, Qian

Advisor dc:contributor.advisor

• Hu, Fei

Contributors dc:contributor

• Kumar, Sunil
• Hong, Xiaoyan
• Song, Aijun
• Freeborn, Todd

• Electrical engineering

Statement dc:rights

• All rights reserved by the author unless otherwise indicated.

Language dc:language.iso

en_US, English

Dc Identifier Other

u0015_0000001_0003467
Mao_alatus_0004D_13967

OAI identifier oai:identifier

oai:ir.ua.edu:123456789/6524