To enhance the accuracy of excavator external forces estimation, a data-physics hybrid-driven excavator external forces estimation method is proposed. MultiLayer Perceptron (MLP) model is applied for nonlinear component modeling and is used to compensate for Rigid Body Dynamics (RBD) models. Gaussian Process Regression (GPR) model is developed to estimate the external forces and uncertainties based on the joint angles, angular velocities, cylinder actuation forces, and actuation end velocities of the excavator. The external forces estimated by the GPR models are used as a new measurement in the Disturbance Kalman Filter (DKF), and the uncertainty is incorporated into the covariance of the process noise. The constructed GPR-DKF estimator is validated on a scaled-down excavator. The results demonstrate the robustness and effectiveness of the proposed method in estimating out-of-distribution samples.

中文翻译：

---

数据物理混合驱动的挖掘机外力估计方法


为了提高挖掘机外力估计的准确性，提出了一种数据-物理混合驱动的挖掘机外力估计方法。多层感知器 (MLP) 模型应用于非线性组件建模，并用于补偿刚体动力学 (RBD) 模型。高斯过程回归 (GPR) 模型用于根据挖掘机的关节角度、角速度、油缸驱动力和驱动端速度来估计外力和不确定性。由探地雷达模型估计的外力被用作干扰卡尔曼滤波器（DKF）中的新测量，并且不确定性被纳入过程噪声的协方差中。构建的 GPR-DKF 估计器在小型挖掘机上进行了验证。结果证明了该方法在估计分布外样本方面的稳健性和有效性。