Neural network-based variable stiffness impedance control for internal/external forces tracking of dual-arm manipulators under uncertainties

The desired interaction between manipulators, objects, and environments has resulted in the internal/external force control for dual-arm manipulators being in increasing demand. Consequently, this study focused on the internal/external force tracking for dual-arm manipulator systems under external disturbances, geometries, and stiffness uncertainties which continuously lead to unsatisfactory internal force tracking. The proposed scheme is based on a two-level adaptive impedance control scheme, where the stiffness coefficient is adjusted to adapt to uncalibrated objects. An object-level hybrid impedance controller was used to regulate the external disturbance to produce a compliant response. A manipulator-level neural network-based variable stiffness impedance controller (NNVSIC) was proposed to regulate the internal force under various uncertainties. Additionally, an adaptive wavelet neural network was designed to compensate for the geometric estimation errors of the object. The variable stiffness coefficient could automatically adapt to an unknown object during the cooperation process. One advantage of the proposed method is that no prior knowledge was required. The same controller parameters could be adapted to various objects. The asymptotic stability of the proposed NNVSIC was proven via Lyapunov stability analysis. A series of experiments were conducted using two self-developed nine-degrees-of-freedom redundant manipulators. Furthermore, hard and soft objects of various geometries and stiffnesses were used to verify the effectiveness of the algorithm. The experimental results demonstrated the efficiency and superiority of the proposed controller through performance comparison with various algorithms.