An Improved Quantum-Behaved Particle Swarm Optimization Algorithm Combined with Reinforcement Learning for AUV Path Planning

In order to solve the problem of fast path planning and effective obstacle avoidance for autonomous underwater vehicles (AUVs) in two-dimensional underwater environment, a path planning algorithm based on deep Q-network and Quantum particle swarm optimization (DQN-QPSO) was proposed. Five actions ar...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Journal of Robotics Ročník 2023; s. 1 - 11
Hlavní autoři: Zhang, HanBin, Shi, XianPeng
Médium: Journal Article
Jazyk:angličtina
Vydáno: New York Hindawi 05.04.2023
John Wiley & Sons, Inc
Wiley
Témata:
Algorithms
Artificial intelligence
Autonomous underwater vehicles
Design
Energy consumption
Machine learning
Mathematical optimization
Neural networks
Obstacle avoidance
Ocean currents
Optimization
Particle swarm optimization
Path planning
Planning
Random variables
Remote submersibles
Simulation
Vortices
ISSN:1687-9600, 1687-9619
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:In order to solve the problem of fast path planning and effective obstacle avoidance for autonomous underwater vehicles (AUVs) in two-dimensional underwater environment, a path planning algorithm based on deep Q-network and Quantum particle swarm optimization (DQN-QPSO) was proposed. Five actions are defined first: normal, exploration, particle explode, random mutation, and fine-tuning operation. After that, the five actions are selected by DQN decision thinking, and the position information of particles is dynamically updated in each iteration according to the selected actions. Finally, considering the complexity of underwater environment, the fitness function is designed, and the route length, deflection angle, and the influence of ocean current are considered comprehensively, so that the algorithm can find the solution path with the shortest energy consumption in underwater environment. Experimental results show that DQN-QPSO algorithm is an effective algorithm, and its performance is better than traditional methods.
Bibliografie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1687-9600
1687-9619
DOI:10.1155/2023/8821906