Ibrahim Sulaiman Mohammed, Awwal Aliyu M, Malik Maulana, Khalid Ruzelan, Benjamin Aida Mauziah, Mohd Nawawi Mohd Kamal, Madi Elissa Nadia
School of Quantitative Sciences, Universiti Utara Malaysia, Sintok, Kedah, Malaysia.
Faculty of Arts and Education, Sohar University, Sohar, Oman.
PeerJ Comput Sci. 2025 May 23;11:e2783. doi: 10.7717/peerj-cs.2783. eCollection 2025.
This study presents a novel gradient-based algorithm designed to enhance the performance of optimization models, particularly in computer science applications such as image restoration and robotic motion control. The proposed algorithm introduces a modified conjugate gradient (CG) method, ensuring the CG coefficient, β κ, remains integral to the search direction, thereby maintaining the descent property under appropriate line search conditions. Leveraging the strong Wolfe conditions and assuming Lipschitz continuity, we establish the global convergence of the algorithm. Computational experiments demonstrate the algorithm's superior performance across a range of test problems, including its ability to restore corrupted images with high precision and effectively manage motion control in a 3DOF robotic arm model. These results underscore the algorithm's potential in addressing key challenges in image processing and robotics.
本研究提出了一种基于梯度的新型算法，旨在提高优化模型的性能，特别是在图像恢复和机器人运动控制等计算机科学应用中。所提出的算法引入了一种改进的共轭梯度（CG）方法，确保CG系数βκ在搜索方向中保持不可或缺，从而在适当的线搜索条件下保持下降特性。利用强 Wolfe 条件并假设Lipschitz连续性，我们建立了该算法的全局收敛性。计算实验证明了该算法在一系列测试问题上的卓越性能，包括其高精度恢复受损图像的能力以及在3自由度机器人手臂模型中有效管理运动控制的能力。这些结果强调了该算法在解决图像处理和机器人技术中的关键挑战方面的潜力。
