Coupled adaptive neural network control for active suspension of deep-sea mining tracked vehicles under hydrodynamic effects and multiple parametric uncertainties

Abstract

Deep-sea mining tracked vehicles are subject to substantial parametric uncertainty from hydrodynamic effects, soft-seabed interaction, suspension nonlinearities, and actuator degradation. This paper proposes a coupled adaptive neural network (CANN) strategy for an 8-DOF six-wheel half-vehicle active suspension. Unlike channel-wise suspension controllers, the method reconstructs a two-dimensional body-level force-tracking error from bounce and pitch accelerations. A nominal backstepping law supplies the baseline generalized force, while a bounded-feature neural estimator with a composite filtering law learns the lumped body disturbance and uncertain equivalent input gain online without requiring numerical uncertainty bounds. A minimum-norm pseudoinverse maps the two-dimensional compensation command to the six suspension actuators. Lyapunov analysis establishes uniform ultimate boundedness of the interconnected tracking, body-force, and weight-estimation errors under the stated assumptions. The resulting framework provides a dimensionally consistent solution to coupled uncertainty compensation and redundant force allocation in deep-sea tracked-vehicle suspension control.