一种新的角加速度估计方法及其应用

首先提出了一种新的基于卡尔曼滤波及牛顿预测的角加速度估计方法,在已知电机驱动系统位置信息的情况下,利用卡尔曼滤波实时估计系统的角加速度;同时采用牛顿预测方法解决估计算法的滞后问题,进一步提高了估计加速度的响应频带．以此为基础,本文进一步分析了利用估计加速度进行反馈控制以增强系统对外扰动的鲁棒性问题,提出了加速度反馈控制策略的设计准则并分析了稳定性．在一个直接驱动机器人关节上针对上述加速度估计及控制方法进行了实验研究:将估计加速度的实验结果与实测加速度(利用加速度计)的实验结果进行了比较分析,从而定量地揭示出估计加速度及其反馈控制在实际系统中的可行性及有效性．

A Newton predictor (NP) is incorporated into the Kalman filter (KF) for online angular acceleration estimation. It intends to reduce the phase lag due to filtering while maintaining or even improving the prediction performance. The acceleration feedback control is applied to design an acceleration close-loop in terms of stability and robustness. Extensive experiments are also conducted on the first joint of a 2-DOF direct-dirve manipulator. Results are compared to those obtained by KF-only estimator and by accelerometer, to demonstrate the improvements of the Kalman filter with Newton predictor (KFNP), as well as the feasibility and validity while the estimated acceleration is used for control in place of the measured acceleration by accelerometer.