Abstract:
Operating unmanned aerial vehicles (UAVs) for data collection is a promising
approach across various practical domains, offering flexibility in challenging environments
to facilitate data collection within sensor networks (SNs). However,
UAV-assisted data collection in SNs faces several challenges, primarily due to
energy constraints at both UAV and SN nodes and the inefficiencies caused by
collecting redundant data. Addressing these issues is crucial for improving the
efficiency of UAV-assisted data collection. Considering the value of information
(VoI) urges the collection of the newly generated data and gives less importance
for collecting old data values. Moreover, the collection of all data may lead to
collect data representing redundant information which may reduce the network efficiency.
This study aims to reduce redundant data collection while deploying the
minimum number of UAVs, minimizing energy consumption and maximizing VoI.
We first formulate the general problem and solve it as a multi-objective optimization
problem. We then decompose the problem into two sub-problems where wepropose real-time approaches including (1) data redundancy avoidance and VoI
evaluation, and (2) dynamic UAV deployment and position adaptation. In the
first problem, the proposed approach clusters SNs and prioritizes non-redundant
data by assigning VoI, while neglecting redundant data. In the second, we consider
optimized UAV position adaptation where we generated the problem as a
multi-objective optimization problem and solved it as a mixed-integer linear programming
problem with constraints related to UAV range, UAV steps, and time
constraints. To address these objectives, our proposed approach incorporates
deep reinforcement learning (RL-DQN) techniques to optimize UAV deployment,
minimizing the number of UAVs while maximizing the number of successfully
collected SNs with non-redundant data. The model considers VoI and energy
constraints of the SNs, enhancing both efficiency and sustainability. The proposed
approach outperforms other algorithms, demonstrating higher efficiency in
terms of UAV deployment, served SN, VoI and energy consumption.
