Adaptive Safety Control for Constrained Uncertain Robotic Systems: A Neural Network-Based High-Order Control Barrier Function Approach

In this paper, a neural network-based high-order control barrier function (NNHoCBF) is proposed to address the safety control problem of constrained uncertain robotic systems, where the radial basis function-based neural network is introduced to reconstruct uncertain robotic systems. By proving the non-negative nature of NNHoCBFs, Lyapunov-like conditions are obtained to ensure the constraint satisfaction of uncertain robotic systems. Moreover, to ensure the safe tracking control for constrained uncertain robotic systems, a minimum energy quadratic program (QP) with Lyapunov-like conditions is constructed as constraints on nominal control inputs, and the safe tracking controllers of robotic systems are then obtained by solving the minimum energy QP. Consequently, the safety and tracking performances of constrained uncertain robotic systems can be guaranteed simultaneously. Finally, simulation tests on a two-link robotic system are conducted to verify the effectiveness of the proposed controller.