Optimal Task Offloading and Trajectory Planning Algorithms for Collaborative Video Analytics With UAV-Assisted Edge in Disaster Rescue

Researchers have designed UAV-enabled edge servers (UES) to assist smart cameras (ECs) and optimize video analytics performance. However, most prior research neglects the impact of ECs' battery constraint on the system performance, especially in disaster scenarios. In this study, we introduce a...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:IEEE transactions on vehicular technology Ročník 73; číslo 5; s. 6811 - 6828
Hlavní autoři: Sun, Hui, Zhang, Xiuye, Zhang, Bo, Sha, Kewei, Shi, Weisong
Médium: Journal Article
Jazyk:angličtina
Vydáno: New York IEEE 01.05.2024
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Témata:
Algorithms
Autonomous aerial vehicles
Batteries
Computation offloading
Deep learning
Disasters
Discrete time systems
double deep q-learning
Edge computing
Energy conservation
Energy consumption
Evolutionary algorithms
Flight
Hovering
Machine learning
Markov processes
Optimization
Q-learning
Resource management
Scheduling
Task analysis
task scheduling
Trajectory
Trajectory planning
Unmanned aerial vehicles
ISSN:0018-9545, 1939-9359
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:Researchers have designed UAV-enabled edge servers (UES) to assist smart cameras (ECs) and optimize video analytics performance. However, most prior research neglects the impact of ECs' battery constraint on the system performance, especially in disaster scenarios. In this study, we introduce a novel time-slot-based UES-assisted system that aims to conserve energy and extend the overall lifetime of EC networks. Our proposed system operates in discrete time slots, wherein the UES alternates between hovering and serving ECs, or flying to a new position within its coverage area to provide better assistance. To minimize the computational overhead of ECs during hovering and serving within a time slot, we present a new task offloading scheme based on the differential evolutionary algorithm. For optimizing the UES's flying trajectory, we formulate the UES movement as a Markov decision process, taking into account system state changes. Subsequently, we design an efficient UES trajectory planning algorithm using double deep Q-learning. This algorithm optimizes energy consumption, feedback reward, and the overall system overhead, resulting in a doubling of the system's lifetime. Simulation results demonstrate that our proposed task offloading algorithm exhibits high accuracy and fast convergence compared to four other state-of-the-art strategies. Moreover, the UES trajectory planning algorithm doubles the system's lifetime while reducing energy consumption and total system overhead.
Bibliografie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0018-9545
1939-9359
DOI:10.1109/TVT.2023.3344281