Adaptive broad learning neural network for integrated guidance and control of maneuvering targets

This paper addresses the integrated guidance and control (IGC) problem for interceptors engaging maneuvering targets under terminal line-of-sight (LOS) angle constraints, model uncertainties, and external disturbances. To estimate unknown target maneuvers within a prescribed time, a prescribed-time extended state observer (PTESO) is developed, in which a trigonometric time-scaling function is introduced to suppress transient peaking and enhance estimation robustness. To handle strong nonlinear uncertainties, an adaptive broad learning neural network (BLNN) with online node growth is proposed to achieve online structural adaptation and improved approximation capability while maintaining a compact network structure. Moreover, a smooth switching mechanism is introduced to coordinate neural approximation with adaptive robust compensation, thereby improving disturbance rejection performance under varying operating conditions. Furthermore, an adaptive command-filter-based backstepping IGC scheme is developed within a logarithmic Lyapunov framework. Unlike conventional quadratic Lyapunov designs, the proposed logarithmic formulation inherently yields bounded tanh ( · ) feedback terms, preventing the excessive growth of recursive control signals. Meanwhile, the adaptive command filter eliminates the explosion of complexity from repeated differentiation while relaxing the requirement for prior knowledge of virtual control derivative bounds. Rigorous stability analysis shows that all closed-loop signals are semi-globally uniformly ultimately bounded (SGUUB). Simulation results demonstrate effective interception and satisfactory terminal LOS angle tracking performance, while Monte Carlo simulations further confirm strong robustness against model uncertainties and external disturbances.