Energy‐efficient animal tracking with multi‐unmanned aerial vehicle path planning using reinforcement learning and wireless sensor networks

In recent years, the integration of wireless sensor networks (WSNs) and unmanned aerial vehicles (UAVs) has been a popular research domain in animal tracking due to animal habitats in harsh environments. A well‐designed WSN network provides a robust mechanism to detect the animal appearances which c...

Full description

Saved in:
Bibliographic Details
Published in:Concurrency and computation 2023-02, Vol.35 (4), p.n/a
Main Authors: Ergunsah, Senol, Tümen, Vedat, Kosunalp, Selahattin, Demir, Kubilay
Format: Article
Language:English
Subjects:
Algorithms
Machine learning
Messages
path planning
reinforcement learning
Topology
Tracking systems
Trajectory planning
UAV
Unmanned aerial vehicles
wild animal tracking
Wireless networks
Wireless sensor networks
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In recent years, the integration of wireless sensor networks (WSNs) and unmanned aerial vehicles (UAVs) has been a popular research domain in animal tracking due to animal habitats in harsh environments. A well‐designed WSN network provides a robust mechanism to detect the animal appearances which can be then collected by UAVs. In this paper, we first design a WSN‐based network model that eliminates the deficiencies of traditional WSNs by intelligently building a WSN topology. Second, we propose a multi‐agent (UAVs) Q‐learning‐based trajectory planning algorithm, which enables UAVs to acquire the detection information of the wild animal on time. Q‐learning algorithm coordinates the UAVs to visit WSN nodes containing fresh animal detection data in a timely manner according to past experience. The experimental results denotes that the proposed animal tracking system delivers significantly higher timely animal detection messages of 45% than the existing algorithms and the single agent Q‐learning approaches. The results also show that since the Zebras' movement contains some randomness, relatively high epsilon value (0.2) in the epsilon‐greedy mechanism of the Q‐learning provides highest rewards (timely delivered messages) for agents (UAVs). In the light of the results, it is clearly seen that intelligently designed WSN and UAV swarm combination could enable tracking of the wild animal in near real time.
ISSN:1532-0626
1532-0634
DOI:10.1002/cpe.7527