Artificial Intelligence for Unmanned Aerial Vehicles (eBook)

Sensing, Communication, and Computing

Shuyan Hu, Kai Li, Xin Yuan, Wei Ni, Xin Wang (Autoren)

eBook Download: EPUB

2026
573 Seiten
Wiley-IEEE Press (Verlag)
978-1-394-36946-1 (ISBN)

Infos

In-depth exploration of machine learning techniques applied to UAV operations and communications, highlighting areas of potential growth and research gaps

Artificial Intelligence for Unmanned Aerial Vehicles provides a comprehensive overview of machine learning (ML) techniques used in unmanned aerial vehicle (UAV) operations, communications, sensing, and computing. It emphasizes key components of UAV activity to which ML can significantly contribute including perception and feature extraction, feature interpretation and regeneration, trajectory and mission planning, and aerodynamic control and operation.

The book considers the notion of security in the UAV network primarily in terms of its underlying rationale. This book also includes a detailed analysis of UAV behavior with respect to time and explores online machine learning-based solutions for UAV-assisted IoT networks.

Additional topics include:

• Joint cruise control and data collection
• Resilience in an AI-aided UAV network against multiple attacks, introducing a flexible and adaptive threshold to alleviate malicious conduct
• Quantification of influencing attributes, quantification of weights affiliated with these attributes, and movement tracking of malicious UAVs
• Integration of contextual information, threshold definitions, and time-variant behavior analysis

Artificial Intelligence for Unmanned Aerial Vehicles is an essential up-to-date reference on the subject for researchers, professors, graduate and senior undergraduate students, and industry professionals in the field.

SHUYAN HU, PHD, is an Associate Professor with the College of Electronics and Information Engineering at Tongji University, Shanghai, China.

XIN YUAN, PHD, is a Senior Research Scientist at CSIRO and an Adjunct Senior Lecturer at the University of New South Wales, Sydney, NSW, Australia.

KAI LI, PHD, serves as a Visiting Research Scientist with the School of Electrical Engineering and Computer Science, TU Berlin, Germany, and is also a Senior Research Scientist with Real-Time and Embedded Computing Systems Research Centre (CISTER), Porto, Portugal.

WEI NI, PHD, is a Senior Principal Research Scientist at CSIRO and a Conjoint Professor at the University of New South Wales, Sydney, NSW, Australia.

XIN WANG, PHD, is a Professor with the College of Future Information Technology at Fudan University, Shanghai, China.

Textauszug

List of Figures

2.1	Illustration on a pair of autonomous UAVs flying random 3D trajectories with smooth turns in a 3D sphere.

2.2	A UAV swarm flies autonomously within a 3D sphere in the presence of external interference from the ground, where is the distance from the ground interferer to the bottom of the 3D sphere, and is the elevation angle from the ground interferer to the UAV in the swarm.

2.3	Two UAVs in the spherical region (Special case: UAV 1 is located on the surface of the 3D spherical region).

2.4	Two UAVs in a 2D disk (Special case: UAV 1 is located on the boundary of the disk).

2.5	The ergodic capacity vs. radius, with the Rician factor , .

2.6	The outage capacity vs. radius, with the Rician factor , , and .

2.7	The outage capacity vs. SNR threshold, with the Rician factor , , and .

2.8	The ergodic capacity vs. SNR threshold, with the Rician factor , , and .

2.9	A top view of part of a 3D ST trajectory, where there are 10 waypoints marked by black dots, the trajectory is in red, and the arcs between consecutive waypoints (i.e., the turn centers and radii) are uniquely determined geometrically.

2.10	An example of a 3D ST trajectory with 3000 waypoints. (a) An example of a segment of a 3D ST trajectory of a UAV within a 3D sphere, where there are 3000 waypoints in the segment to keep the figure clear. (b) The projection of the 3D trajectory on the -plane.

2.11	The outage probability of an arbitrary UAV vs. the number of UAVs, , under different Rician factors, where the number of waypoints , , , and .

2.12	The outage probability of an arbitrary UAV vs. the number of UAVs under different SNR/SINR thresholds, where and .

2.13	The outage probability of an arbitrary UAV vs. the transmit power of UAVs for different values of SNR threshold, where , , , , and the Rician factor is .

2.14	The outage probability of an arbitrary UAV vs. the radius of the sphere , under different total numbers of UAVs, where the number of waypoints , the Rician factor , , , and .

2.15	The outage probability in the presence of the ground jammer vs. the height under different Rician factor , where , , , , , and .

2.16	The number of UAVs in the swarm both in the presence and absence of the ground jammer vs. radius under outage probability threshold , where , , , , , and .

3.1	The system of interest, where there is a pair of legitimate transmitter and receiver on the ground, and an aerial eavesdropper flying within the transmission range of the transmitter. The eavesdropper follows the ST mobility model.

3.2	Illustration of the system model, where aerial eavesdroppers fly random ST mobility within a spherical cap with polar angle , .

3.3	The scenario of a bidirectional ground link, where the aerial eavesdropper flies within the overlapping coverage region of both the legitimate ground nodes.

3.4	The ergodic secrecy rate vs. the radius of the eavesdropper’s flight region, , under different values of path loss, , where , and .

3.5	The ergodic secrecy rate vs. , in the presence of an aerial eavesdropper under different values of path loss exponent, , where .

3.6	The -outage secrecy rate vs. the radius of the eavesdropper’s flight region, , under different values of both path loss exponent and Rician factors, where , , and .

3.7	The -outage secrecy rate vs. the outage probability, , in the presence of an aerial eavesdropper flying in 3D spherical spaces, where , , and .

3.8	The -outage secrecy rate vs. , under different values of both target secrecy rate and Rician factor.

3.9	The secrecy rates (Scenario 1) under different values, where , , , and .

3.10	Ergodic secrecy capacity of the eavesdroppers conducting SC and MRC vs. and , where , , and .

4.1	Surface area of the co-centric 3D hemisphere with the radius intersecting the 3D spheres centered at and with radius of the safety distance.

4.2	Ergodic secrecy capacity of aerial eavesdroppers conducting SC and MRC against the minimum altitude , where and .

4.3	Outage secrecy capacity of the eavesdroppers conducting SC and MRC vs. and , where , , , and .

4.4	The density of aerial eavesdroppers vs. and , under both SC and MRC, where , , and . (c) provides the contours for (a), where the values, such as 2 × 10−7, stand for the cut-off density of eavesdroppers (per cubic meter).

4.5	Ergodic secrecy capacity of the ground link vs. in the presence and absence of oscillator phase noises at the aerial eavesdroppers, where , , , the path loss at a reference distance is , and (under , and , and ).

4.6	(a) The ratio of the eavesdropping probability between MRC and SC eavesdropping and (b) the ratio of captured traffic between MRC and SC eavesdropping, where , , and .

5.1	In an aerial EdgeIoT, the UAV carrying a lightweight edge server controls the trajectory and patrols over the area of interest. The IoT devices can offload their computation tasks to the UAV for processing or forwarding to the remote cloud server.

5.2	The combinatorial learning architecture based on GNN and A2C for aerial EdgeIoT.

5.3	TMR in terms of training episodes.

5.4	Number of offloaded tasks and buffered tasks at each IoT device, where and . Each error bar provides the standard deviation of 10 experiments.

5.5

Flight trajectories of the UAV, which is controlled by the proposed GNN-A2C framework, given different numbers of IoT devices, values, distribution patterns of the IoT devices, and training iterations. (a) and ; (b) and ; (c) given uniform distribution; (d) given normal distribution; (e) and training iterations = 1500; (f) and training iterations = 3000.

5.6	TMR vs. the aerial EdgeIoT’s scalability , where is 100. Each error bar provides the standard deviation of 10 independent experiments.

5.7	TMR vs. , where . Each error bar shows the standard deviation of 10 independent experiments.

5.8	Runtime measurement of the GNN-A2C, where the number of training iterations is 50 or 1500.

6.1	A UAV-enabled video surveillance system.

6.2	An illustration of the forces on an ascending aircraft, where , , , and stand for weight, drag, lift, and thrust, respectively.

6.3	An illustration of the online scheme in the 2D space.

6.4	A bird’s view illustration of the online control approach.

6.5	Original value and approximation of solar power.

6.6	Convergence of the original objective values in (6.35a).

6.7	Convergence of the monitor 3D trajectory.

6.8	Relative monitor 2D trajectory observed by the target.

6.9	Objective values under different values of .

6.10	Monitor 3D trajectory under different values of .

6.11	Objective values under different schemes.

6.12	Monitor horizontal trajectory under different schemes.

6.13	Monitor altitude under different schemes.

6.14	Energy saving of the proposed...

Produkt-Details

• Erscheint lt. Verlag: 5.1.2026
• Sprache: englisch
• Themenwelt: Technik ► Elektrotechnik / Energietechnik
• Schlagworte: capacity UAV • communication UAV • Cruise control • cyber threats • joint sensing • Multi-agent Systems • reconfigurable intelligent surfaces • UAV AI • UAV connectivity • UAV Control • UAV privacy • UAV security • UAV surveillance
• ISBN-10: 1-394-36946-8 / 1394369468
• ISBN-13: 978-1-394-36946-1 / 9781394369461