| 孙明轩,吴雨芯,张钰.二次曲线型终态神经网络:时变神经计算与冗余机械臂重复运动规划[J].高技术通讯(中文),2020,30(10):1037~1048 |
| 二次曲线型终态神经网络:时变神经计算与冗余机械臂重复运动规划 |
|
| |
| DOI：doi：10.3772/j.issn.1002-0470.2020.10.006 |
| 中文关键词: 终态神经网络； 时变线性矩阵方程； 冗余机械臂； 重复运动规划 |
| 英文关键词: terminal neural network, time-varying linear matrix equation, redundant manipulator, repeatable motion planning |
| 基金项目: |
|
| 摘要点击次数: 2579 |
| 全文下载次数: 1701 |
| 中文摘要: |
| 提出二次曲线型终态神经网络，包括双曲线型、椭圆型和抛物线型3种终态神经网络，网络各变量取值有限，易于实现。详细分析了这类网络的有限时间收敛特性，给出了具体的收敛时间表达式，并以双曲线型终态神经网络为例，将其应用于时变线性矩阵方程求解以及机器人轨迹规划。对于一般时变线性矩阵方程的求解，讨论了时变Lyapunov方程和时变Sylvester方程，分别给出了终态神经网络用于求解这2类时变矩阵方程的计算结果，验证了该网络能在有限时间内精确收敛到理论解。对于冗余机械臂重复运动规划，文中将重复运动指标设计为终态收敛性能指标，在初始位置偏移的情况下，利用该终态神经网络进行求解，实现冗余机械臂有限时间收敛的重复运动规划任务。 |
| 英文摘要: |
| Conic terminal neural networks are proposed based on the geometric features of the curves, which include three categories: the hyperbolic, elliptic and parabolic types. The variables involved in the networks are all bounded, which makes such networks easily implementable. The finite-time convergence is established, with derivations for the explicit expressions for their settling time functions. The hyperbolic networks are applied to solve the time-variant linear matrix equations and for the motion planning of robotic manipulators, respectively. Lyapunov and Sylvester equations are taken as examples. The finite-time convergence is achieved due to the use of the terminal network, and the resultant solutions are presented with the expected accuracy. By using a terminal criterion, the repeatable motion planning of the manipulator undertaken is realized in the presence of the initial positional shift. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
