极限学习机
计算机科学
人工神经网络
控制(管理)
人工智能
控制工程
控制系统
智能控制
实时计算
反向传播
机器学习
内存管理
钥匙(锁)
前馈神经网络
过程控制
工程类
短时记忆
智能网
实时控制系统
移动机器人
作者
Aochen Ma,Bin Xian,Fei Geng,Kai Xu
标识
DOI:10.1109/tii.2026.3732451
摘要
This article presents a fault-tolerant control (FTC) framework with position-tracking capability for a small-size unmanned helicopter under unknown actuator faults. A sliding surface that satisfies the transient and steady-state performance requirements is constructed based on a new prescribed performance function. Then, a network that integrates the rapid feature extraction capability of the extreme learning machine (ELM) with the temporal memory characteristics of the long short-term memory (LSTM) network is developed to approximate and compensate for unknown actuator faults in real-time. Based on the learning ability of this ELM-LSTM network, an attitude fault-tolerant controller is formulated. Subsequently, a robust position-loop controller is designed to enhance the system's resilience against disturbances and actuator efficiency degradation, ensuring satisfactory position-tracking performance even under faults. A detailed stability analysis proves the stability of the small-size unmanned helicopter system. Finally, real-time hardware-in-the-loop (HIL) experiments on a small-size helicopter platform are conducted to validate the proposed control strategy. Experimental results confirm that the method enables rapid recovery of attitude stability after faults, and maintains high-accuracy position tracking performance.
科研通智能强力驱动
Strongly Powered by AbleSci AI