abstract
- This work describes the design of an identification algorithm for robotic systems, which are represented by a class of nonlinear multi-input-multi-output systems with model uncertainties. The proposed identifier exploits the approximation properties of differential neural networks. The learning or adaptive laws are obtained using barrier Lyapunov functions with an exponentially decreasing barrier. Applying such a function results in a bounded identification error with exponential convergence and predefined decay. Additionally, it ensures the convergence of the weights for the activation functions to the `best fit¿ values. The proposed identifier is applied to a Cartesian robot's virtual model and also to the Kinova Gen3 Lite robot operating in haptic mode. The results show a successful estimation of the states for both robots, one with a predefined path controlled by a proportional-derivative controller and the second from an experiment in haptic mode in which the human user induces the movement. The identification error is bounded by the proposed exponentially decreasing function with a predefined decay for both robots. © 2025 Elsevier Inc.