基于回归神经网络多机械臂运动控制研究

发布时间：2018-05-16 09:07

本文选题：对偶回归神经网络 + 二次优化　；参考：《控制工程》2017年11期

【摘要】：该文将多协作冗余机械臂的运动控制问题转化成一个可完全分解的约束二次求解问题,并将该约束二次求解问题拆分成一系列独立的子问题,每个子问题对应于单个机械臂的运动控制模型,其大大简化了多协作机械臂的运动控制模型。另外,该文采用对偶回归神经网络求解该运动控制模型,每个独立的神经网络核对应于一个机械臂的控制模型。该文首次将对偶回归神经网络用于多冗余机械臂系统的控制问题中,具有很强的应用研究意义。与现有的对偶回归神经网络相比,该文提出的对偶回归神经网络结构非常特殊,其多项性能将大大提高,使得多协作机械臂的运动控制问题变成完全分布式。理论验证结果表明,该文提出的模型是全局稳定的并且可得到全局最优解。仿真实验结果表明本论文提出的方案是有效的。该文的研究结果可为多协作冗余机械臂的控制理论和技术发展提供新的思路。
[Abstract]:In this paper, the motion control problem of multi-cooperative redundant manipulators is transformed into a fully decomposable constraint quadratic solution problem, and the constrained quadratic solution problem is divided into a series of independent sub-problems. Each sub-problem corresponds to the motion control model of a single manipulator, which greatly simplifies the motion control model of a multi-cooperative manipulator. In addition, the dual regression neural network is used to solve the motion control model. Each independent neural network kernel corresponds to the control model of a manipulator. In this paper, the dual regression neural network is first applied to the control of multiple redundant manipulator systems, which has a strong application significance. Compared with the existing dual regression neural networks, the dual regression neural network structure proposed in this paper is very special, and its multi-performance will be greatly improved, which makes the motion control problem of multi-cooperative manipulator become completely distributed. The theoretical results show that the proposed model is globally stable and the global optimal solution can be obtained. The simulation results show that the proposed scheme is effective. The results of this paper can provide new ideas for the development of control theory and technology of multi-cooperative redundant manipulators.
【作者单位】：深圳信息职业技术学院信息技术研究所;江西理工大学机电工程学院;
【基金】：广东省自然科学基金(S2012010008997&2015A030313587) 广东省高等学校优秀青年教师培养计(YQ2013194) 深圳市科技计划(JCYJ20140418100633634&JCYJ20150417094158025&GRCK20170424095924228)
【分类号】：TP183;TP241

【相似文献】