NN-enhanced observer for estimating external forces and adaptive variable impedance control for contact force tracking in uncertain environments

This study investigates the issue of force tracking control when a flexible-joint robotic (FJR) arm interacts with an uncertain environment, aiming to achieve precise desired force tracking in interactions with complex and variable environments. Firstly, a neural network (NN)- enhanced momentum observer is proposed by combining a momentum observer with NN technology, using joint torque sensors to achieve an accurate estimation of external forces on the end-effector of a flexible joint robotic arm, enhancing the accuracy of external force estimation. Then, based on this external force estimation, an adaptive variable impedance control (VIC) method is designed, which dynamically adjusts the damping coefficient of the impedance model in real-time, allowing the robotic arm to adapt to uncertainties in unknown environments; based on this impedance model, a control method is derived that directly generates Cartesian space reference velocities, achieving the desired VIC control and precise force tracking control in interactions with unknown environments. Finally, comparative experiments demonstrate the effectiveness and advantages of this method. The experimental results show that this method significantly improves the accuracy of external force estimation, possesses adaptive force tracking control capability in unknown environments, and enhances the interaction performance of the robotic arm with complex unknown environments.