学术报告

A Proximal Linearized Algorithm for Dual Quaternion Optimization with Applications in Hand-Eye Calibration

陈艳男（华南师范大学）

报告时间：2023年6月30日 星期五  15：00-16：00

报告地点：沙河主楼E404

腾讯会议：802-256-565

报告摘要: Many kinetic problems arising from robotics are modeled in optimization of real function in dual quaternion variables. For example, hand-eye calibration fits data to find dual quaternion variables, which integrate the relative position and orientation between a robot gripper and a camera mounted rigidly on the gripper. But now derivatives of real function in dual quaternion variables are unwieldy. To solve optimization of real function in dual quaternion variables, we utilize eight-dimensional vectors to represent dual quaternions. Then the unit dual quaternion constraint is turned into two quadratic equations. Using optimization theory, we derive an algorithm for computing the projection of dual quaternions onto a unit dual quaternion set, which is closed, nonconvex and unbounded. Based on this projection, we propose a proximal linearized algorithm for optimization of real function in dual quaternion variables. The global convergence of the proximal linearized algorithm is analyzed. Finally, numerical experiments on hand-eye calibration problems and dual quaternion-based regressions illustrate the effectiveness of the proposed proximal linearized algorithm.

报告人简介：陈艳男，博士，华南师范大学太阳成集团tyc7111cc副教授。2013年在南京师范大学获得博士学位，陈博士已发表SCI论文30余篇，代表性论文发表于SIAM J. Matrix Anal. Appl., SIAM J. Sci. Comput., Math. Comput.等国际刊物，参与撰写了一本专著《Tensor Eigenvalues and Their Applications》在Springer出版，完成国家自然科学基金2项，现主持国家自然科学基金面上项目1项，获得2020年度广东省自然科学奖二等奖。

邀请人：崔春风