Neural network-based practical prescribed-time bearing-constrained secure formation control for second-order nonlinear multi-mobile robot systems.

Abstract

This paper investigates bearing-constrained formation control for nonlinear multi-mobile robot systems under false data injection attacks. Radial basis function neural networks are employed to approximate nonlinear dynamics of robots, as well as attacks, aiming to mitigate their adverse impacts. To ensure that multi-mobile robot systems accomplish formation tasks within preset time, a distributed control protocol is designed based on the framework of practical prescribed-time convergence. In unknown environments where GPS signals are denied, onboard sensors are utilized to acquire relative bearings and distances, which also means a reduced dependence on external infrastructure. Finally, considering that the actual communication range limitations may lead to the dynamic changes of available neighbor information, the numerical and physics-based simulations are conducted to evaluate the practicality of the proposed algorithm in the dynamic sensing environment.