一种基于PILCO算法的智能浮体运动控制方法

张 尚; 杨 睿; 陈 震; 黎 明

doi:10.11993/j.issn.2096-3920.2021.05.005

一种基于PILCO算法的智能浮体运动控制方法

doi: 10.11993/j.issn.2096-3920.2021.05.005

张尚¹,
杨睿^1,2,
陈震^1,2,
黎明^1,2

1.
中国海洋大学工程学院, 山东青岛, 266100
2.
山东省海洋智能装备技术工程研究中心, 山东青岛, 266100

基金项目: 国家自然科学基金项目资助(51709245); 国家重点研究发展计划项目资助(2017YFC1405203)

详细信息

作者简介:
张尚(1996-), 男, 在读硕士, 主要研究方向为海上可重构智能浮体控制系统研究.

中图分类号: U674.38 TP242.6 TP181
计量
- 文章访问数: 99
- HTML全文浏览量: 4
- PDF下载量: 26
- 被引次数: 0
出版历程
- 收稿日期: 2020-10-20
- 修回日期: 2020-12-17
- 刊出日期: 2021-10-31

Motion Control Method of Autonomous Surface Vehicle Based on the PILCO Algorithm

1.
College of Engineering, Ocean University of China, Qingdao 266100, China
2.
Shandong Marine Intelligent Equipment Technology Engineering Research Center, Qingdao 266100, China

摘要

摘要: 随着人们对海洋探索的不断深入, 开发一种自主性强、灵活度高、可重构的智能浮体(ASV)至关重要。文中以四推进器ASV为研究对象, 建立了其动力学模型, 基于概率推理的学习控制算法设计了控制器, 并进行了定点控制和轨迹跟踪的仿真实验。仿真结果表明: ASV仅需进行少量的实验即可获得自主学习控制策略, 在有水流扰动或采用近似动力学模型时, 能够实现对其的运动控制, 从而验证了文中算法的有效性。
- 智能浮体 /
- 基于概率推理的学习控制 /
- 定点控制 /
- 轨迹跟踪
Abstract: A highly autonomous, flexible, and reconfigurable autonomous surface vehicle(ASV) must be developed to fulfill the needs for ocean exploration. In this study, an ASV composed of four thrusters is analyzed by establishing the dynamic model of the ASV, designing its controller based on the probabilistic inference learning to control(PILCO) algorithm, and conducting simulation experiments of fixed-point control and trajectory tracking. The simulation results show that the ASV model can autonomously learn the control strategy in a small number of experiments and realize motion control during a water flow disturbance or when using an approximate dynamic model, thereby verifying the effectiveness of the proposed algorithm
- autonomous surface vehicle(ASV) /
- probabilistic inference learning to control(PILCO) /
- fixed-point control /
- trajectory tracking

HTML全文

参考文献(12)

[1]	Park S, Kayacan E, Ratti C, et al. Coordinated Control of a Reconfigurable Multi-vessel Platform: Robust Control Approach[C]//2019 International Conference on Robotics and Automation(ICRA). Montreal, Canada: IEEE, 2019.
[2]	Lu Y, Zhang G, Qiao L, et al. Adaptive Output-feedback Formation Control for Underactuated Surface Vessels[J]. International Journal of Control, 2020, 93(3): 400-409.
[3]	Woo J, Yu C, Kim N. Deep Reinforcement Learning-based Controller for Path Following of an Unmanned Surface Vehicle[J]. Ocean Engineering, 2019, 183: 155-166.
[4]	Paulos J, Eckenstein N, Tosun T, et al. Automated Self-assembly of Large Maritime Structures by a Team of Robotic Boats[J]. IEEE Transactions on Automation Science and Engineering, 2015, 12(3): 958-968.
[5]	Wang W, Mateos L A, Park S, et al. Design, Modeling, and Nonlinear Model Predictive Tracking Control of a Novel Autonomous Surface Vehicle[C]//2018 IEEE International Conference on Robotics and Automation(ICRA). Brisbane, Australia: IEEE, 2018: 6189-6196.
[6]	Mnih V, Kavukcuoglu K, Silver D, et al. Playing Atari with Deep Reinforcement Learning[J]. arXiv, (2013-12-19) [2021-09-01]. https://arxiv.org/abs/1312.5602.
[7]	Deisenroth M, Rasmussen C E. PILCO: A Model-based and Data-efficient Approach to Policy Search[C]// Proceedings of the 28th International Conference on Machine Learning(ICML-11). Bellevue, Washington, USA: ICML, 2011: 465-472.
[8]	Ramirez W A, Leong Z Q, Nguyen H D, et al. Exploration of the Applicability of Probabilistic Inference for Learning Control in Underactuated Autonomous Underwater Vehicles[J]. Autonomous Robots, 2020, 44(6): 1121-1134.
[9]	郭宪. 深入浅出强化学习: 原理入门[M]. 北京: 电子工业出版社, 2018.
[10]	Fossen T I. Guidance and Control of Ocean Vehicles[M]. New Jersey: John Wiley & Sons, 1994.
[11]	陈虹, 刘志远, 解小华. 非线性模型预测控制的现状与问题[J]. 控制与决策, 2001, 16(4): 385-391. Chen Hong, Liu Zhi-yuan, Xie Xiao-hua. Nonlinear Model Predictive Control: The State and Open Problems[J]. Control and Decision, 2001, 16(4): 385-391.
[12]	Lillicrap T P, Hunt J J, Pritzel A, et al. Continuous Control with Deep Reinforcement Learning[EB/OL]. ArXiv, (2015 -09-01) [2021-09-01]. https://www.researchgate.net/publ- ication/281670459_Continuous_control_with_deep_rein- forcement_learning.