Scholarlink

<p>Effective control of rubber-tired gantry cranes (RTGCs) is essential for optimizing seaport operations, especially as increasing container traffic demands greater efficiency. This paper presents an advanced nonlinear Lyapunov control approach for RTGCs to address uncertainties in load transfer operations, such as variations in payload mass and cable length. To manage these uncertainties, k-means clustering is used to group them into distinct clusters, with the optimal control strategy for each cluster determined using Bayesian optimization, specifically Gaussian process with Markov chain Monte Carlo (GP-MCMC). This method outperforms other techniques, including Gaussian process (GP), entropy search, and Bohamia neural network, as validated through 1000 randomized variations in mass and cable length, demonstrating superior results in terms of mean cost, maximum error, and standard deviation. To further address uncertainties, a neural network is employed to predict the active uncertainty cluster based on system states, enabling adaptive switching between controllers. The proposed neural-based Lyapunov switching control demonstrates significant improvements over Lyapunov optimal control methods that lack switching capabilities, as evidenced by extensive Monte Carlo simulation results. Specifically, the proposed method achieves over an 11% reduction in average absolute error, a 34% reduction in standard deviation, a 20% reduction in the 95th percentile of error, and a 46% reduction in maximum error.</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

Enhancing RTGC nonlinear control using neural switching lyapunov for robust payload transfer under uncertainty

  • 期刊论文

  • 发表日期：

    2025年9月27日

查看全文
  • Steven Bandong,
  • Yul Yunazwin Nazaruddin,
  • Endra Joelianto

摘要

Effective control of rubber-tired gantry cranes (RTGCs) is essential for optimizing seaport operations, especially as increasing container traffic demands greater efficiency. This paper presents an advanced nonlinear Lyapunov control approach for RTGCs to address uncertainties in load transfer operations, such as variations in payload mass and cable length. To manage these uncertainties, k-means clustering is used to group them into distinct clusters, with the optimal control strategy for each cluster determined using Bayesian optimization, specifically Gaussian process with Markov chain Monte Carlo (GP-MCMC). This method outperforms other techniques, including Gaussian process (GP), entropy search, and Bohamia neural network, as validated through 1000 randomized variations in mass and cable length, demonstrating superior results in terms of mean cost, maximum error, and standard deviation. To further address uncertainties, a neural network is employed to predict the active uncertainty cluster based on system states, enabling adaptive switching between controllers. The proposed neural-based Lyapunov switching control demonstrates significant improvements over Lyapunov optimal control methods that lack switching capabilities, as evidenced by extensive Monte Carlo simulation results. Specifically, the proposed method achieves over an 11% reduction in average absolute error, a 34% reduction in standard deviation, a 20% reduction in the 95th percentile of error, and a 46% reduction in maximum error.