基于离线学习输入整形的柔性关节机械臂振动控制研究

Vibration control of a flexible joint manipulator based on off-line learning input shaping

输入整形技术是抑制柔性关节机械臂残余振动的一种有效控制策略,然而传统的输入整形器,在设计时需要依赖系统精确的数学模型。针对这一问题,提出了一种新的两脉冲输入整形器参数学习策略;通过振动幅值测量,利用寻优方法(外推内插法和黄金分割法)离线调整整形器参数,不需要精确的数学模型即可找到使残余振动幅值最小的最优参数。为了验证所提方法的有效性,建立了柔性关节机械臂实验平台,进行了基于离线学习输入整形器的柔性关节机械臂振动控制实验。实验结果验证了所提方法的有效性和可行性。

The input shaping technique（ IST） is an effective control strategy to suppress residual vibration of a flexible joint manipulator.However,an exact model is essential for designing a traditional input shaper.Aiming at this problem,a new learning strategy for parameters of a two-impulse input shaper was proposed.The shaper＇s parameters were off-line updated with the optimization methods including the extrapolation and interpolation method,and the golden section one through vibration amplitude measurement.The optimal parameters to minimize residual vibration amplitude were found without an exact model.In order to verify the effectiveness of the proposed method,a test platform for flexible joint manipulators was built,and the tests for residual vibration suppressing of a flexible joint manipulator using off-line learning input shaping were conducted.The test results verified the validity and feasibility of the proposed method.