Optimized neural-network-assisted model-based motion control of hydraulic manipulators under unmodeled uncertainties

Multi-degree-of-freedom (multi-DoF) hydraulic manipulators are widely used in heavy-duty tasks due to their high power density. However, their inherent nonlinearities and uncertain dynamics present significant challenges to precise control. Traditional model-based control methods, while effective in addressing these nonlinearities, rely heavily on accurate dynamic models, which complicates controller design and increases computational demands. To address these limitations, this paper proposes a neural-network-assisted adaptive robust control approach that reduces the dependence on precise modeling by compensating for unmodeled dynamics. Radial basis function neural networks (RBFNNs) are employed to approximate the uncertain dynamics, with the network structure optimized using the K-means++ algorithm to enhance approximation accuracy and computational efficiency. Additionally, desired signals are used instead of measured signals to mitigate sensitivity to measurement noise. The proposed method is rigorously analyzed to guarantee the stability and asymptotic tracking performance of the closed-loop system. Experimental results on a hydraulic manipulator validate the effectiveness of the approach, demonstrating substantial improvements in control performance compared with conventional model-based strategies.