Scholarlink

<p>Real-time inverse kinematics (IK) for robotic manipulators involves solving non-convex optimization problems subject to strict joint limits and singularities. Standard numerical solvers rely on local linearization and are inherently sensitive to initial conditions, often resulting in stagnation in local minima. Concurrency addresses this limitation by enabling the simultaneous exploration of the configuration space using diverse initialization strategies. While existing concurrent solvers have shown promise, this paper introduces Concurrent Inverse Kinematics Solving (CoIKS), a scalable and modular framework that generalizes the concurrent approach. CoIKS leverages a multi-threaded architecture to execute an arbitrary number of independent solvers in parallel, enabling flexible and highly scalable search strategies tailored to modern multi-core processors. The framework supports operational modes optimized for speed, joint-space distance, or manipulability. We validate CoIKS through extensive evaluations on 15 distinct manipulators. The results, from over 10,000 randomly generated poses and a continuous helical path-following benchmark, confirm that this concurrent approach effectively mitigates stagnation in local minima. Our scalable architecture consistently achieves higher success rates and significantly faster computation times than established concurrent solvers such as TRAC-IK. The findings highlight the benefits of a general-purpose, scalable concurrent framework for developing robust, high-performance IK solutions. The code for the framework and the benchmarks presented in this paper are publicly available at <a href="https://github.com/jcolan/CoIKS">https://github.com/jcolan/CoIKS</a>.</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

Optimizing Inverse Kinematics for Robotic Manipulators Through Concurrent Solving

  • 期刊论文

  • 发表日期：

    2026年5月20日

查看全文
  • Jacinto Colan,
  • Ana Davila,
  • Yasuhisa Hasegawa

摘要

Real-time inverse kinematics (IK) for robotic manipulators involves solving non-convex optimization problems subject to strict joint limits and singularities. Standard numerical solvers rely on local linearization and are inherently sensitive to initial conditions, often resulting in stagnation in local minima. Concurrency addresses this limitation by enabling the simultaneous exploration of the configuration space using diverse initialization strategies. While existing concurrent solvers have shown promise, this paper introduces Concurrent Inverse Kinematics Solving (CoIKS), a scalable and modular framework that generalizes the concurrent approach. CoIKS leverages a multi-threaded architecture to execute an arbitrary number of independent solvers in parallel, enabling flexible and highly scalable search strategies tailored to modern multi-core processors. The framework supports operational modes optimized for speed, joint-space distance, or manipulability. We validate CoIKS through extensive evaluations on 15 distinct manipulators. The results, from over 10,000 randomly generated poses and a continuous helical path-following benchmark, confirm that this concurrent approach effectively mitigates stagnation in local minima. Our scalable architecture consistently achieves higher success rates and significantly faster computation times than established concurrent solvers such as TRAC-IK. The findings highlight the benefits of a general-purpose, scalable concurrent framework for developing robust, high-performance IK solutions. The code for the framework and the benchmarks presented in this paper are publicly available at https://github.com/jcolan/CoIKS.