Output-Feedback Finite-Time Safety-Critical Coordinated Control of Path-Guided Marine Surface Vehicles Based on Neurodynamic Optimization

In the presence of static and moving obstacles, this article investigates an output-feedback finite-time safety-critical coordinated control method of multiple under-actuated marine surface vehicles (MSVs) subject to velocity and input constraints. Specifically, based on robust exact differentiators, a finite-time state observer (FTSO) is first developed to recover the unavailable velocities while estimating the total disturbances containing model uncertainties and environmental disturbances. Next, with the aid of estimated velocities from FTSO, a nominal finite-time guidance law is designed for achieving the distributed formation of MSVs at the kinematic level. By the forward invariance principle, finite-time control barrier functions (FTCBFs) are used to construct the collision-free velocity sets for the multi-MSV system. To unify the control and safety objectives, quadratic optimization problems are formulated under collision-free velocity sets and velocity constraints. To facilitate real-time implementations, one-layer recurrent neural networks are employed to solve the quadratic optimization problem. Then, a nominal finite-time control law based on FTSO is presented at the kinetic level. The optimal control laws are solved within the input constraints. All error signals of the closed-loop system are proved to be uniformly ultimately bounded, and the distributed formation of multiple MSVs is ensured to be safe. Simulation results are provided to demonstrate the effectiveness and superiority of the proposed FTCBF-based method.

Automated summary

This article investigates a coordinated control method for multiple under-actuated marine surface vehicles in the presence of static and moving obstacles. The method includes a finite-time state observer and a finite-time guidance law for distributed formation. Control barrier functions are used to construct collision-free velocity sets, and quadratic optimization problems are formulated to unify control and safety objectives. Recurrent neural networks are employed for real-time implementation, and a finite-time control law is presented at the kinetic level. Simulation results demonstrate the effectiveness of the proposed method.