Scholarlink

<p>A complex and unstable device that simulates real-world systems like cranes and rocket stabilization, a fourth-order nonlinear open-loop unstable dynamical system (FNO-UDS) is known as a rotary inverted pendulum (RIP), developing stable control solutions for this intricate and unstable system elastic coupling and control strategy is a challenge set for the researcher. A common tool for evaluating the efficacy of recently created control algorithms is RIP, FNO-UDS. This work investigates the open loop state-feedback type of stabilization control for RIP systems, as well as the simulation, dynamic modeling, and RIP systems’ swing-up and stabilization control. The stability of an RIP system is managed by a proportional integral derivative controller tuned convolutional neural network (PID tuned CNN). The PID controller adjusts the optimal control actions after the CNN has learned the mapping between the system states. The proposed model’s tuned and returned state was evaluated after Mean Square Error (MSE) of the various activation functions in the CNN was analyzed for better enhancement. To achieve the intended performance metrics, data is gathered, trained, and blended. The comparative results show that the method offers superior control effectiveness regarding getting enough short-term, stable, and reliable replies from a certain RIP system.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Enhanced Stability and Control of Rotary Inverted Pendulum Systems Using Deep Learning Control Approach

  • 期刊论文

  • 发表日期：

    2025年7月15日

查看全文
  • Qianmai Peng

摘要

A complex and unstable device that simulates real-world systems like cranes and rocket stabilization, a fourth-order nonlinear open-loop unstable dynamical system (FNO-UDS) is known as a rotary inverted pendulum (RIP), developing stable control solutions for this intricate and unstable system elastic coupling and control strategy is a challenge set for the researcher. A common tool for evaluating the efficacy of recently created control algorithms is RIP, FNO-UDS. This work investigates the open loop state-feedback type of stabilization control for RIP systems, as well as the simulation, dynamic modeling, and RIP systems’ swing-up and stabilization control. The stability of an RIP system is managed by a proportional integral derivative controller tuned convolutional neural network (PID tuned CNN). The PID controller adjusts the optimal control actions after the CNN has learned the mapping between the system states. The proposed model’s tuned and returned state was evaluated after Mean Square Error (MSE) of the various activation functions in the CNN was analyzed for better enhancement. To achieve the intended performance metrics, data is gathered, trained, and blended. The comparative results show that the method offers superior control effectiveness regarding getting enough short-term, stable, and reliable replies from a certain RIP system.