Unmanned Aerial Vehicles (UAVs) have been widely applied in the field of Search and Rescue (SAR). Due to flexibility, UAVs are capable of rapidly deployment in disaster scenarios. To improve efficiency, multiple UAVs form a team to establish networks and collaboratively perform target search tasks. However, existing studies ignore network interference issues that limit multi-UAV information sharing and affect target search efficiency. In addition, inefficient searching by UAVs leads to excessive consumption of flight energy. Thus, in this paper, to enhance efficiency, we propose a method based on Deep Reinforcement Learning (DRL) to make decisions for UAVs. To avoid information redundancy, Spatial Information Entropy (SIE) is utilized to characterize the information density of multi-UAV. To expand the search range of UAVs, maximum-entropy RL (MERL) is utilized to obtain augmented rewards. To cope with the limited network, fuzzy logic-based decision-making rules are proposed for UAVs with different communication statuses. To simultaneously optimize target coverage and energy consumption of UAVs, the Fuzzy-DRL algorithm is proposed based on Multi-Agent Deep Deterministic Policy Gradient (MADDPG) algorithm. Simulation results indicate that the Fuzzy-MADDPG algorithm can optimize the average rewards of UAVs, improve the target coverage rate, and reduce energy consumption.
Fuzzy-MADDPG based Multi-UAV Cooperative Search in Network-limited Environments
19.11.2024
1750886 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Design of a Multi-unmanned Aerial Vehicle Cooperative Hunting System Based on MADDPG
Springer Verlag | 2024
|Research on multi-UAV cooperative pursuit and confrontation strategy based on P3C-MADDPG algorithm
DOAJ | 2023
|Multi-vehicle joint longitudinal control method based on improved MADDPG
Europäisches Patentamt | 2023
|