Three-dimensional self-learning optimal collaborative guidance strategy for multi-missile salvo attacks on variable maneuvering target

This paper delves into an optimal collaborative guidance strategy for salvo attacks on variable maneuvering target. The objective is to conserve energy consumption while guaranteeing the coordinated combat performance of multi-missile system. Based on the three-dimensional (3D) engagement nonlinear dynamic model, the collaborative guidance law with line-of-sight (LOS) angle and time constraints is decomposed into three components. In the LOS direction, finite time simultaneous attack is achieved by exploiting the multi-agent consensus theory and a dynamic integral sliding mode control (SMC). This eliminates the necessity for accurate time-to-go estimation. In the normal directions of LOS, finite-time tracking of the desired LOS angle is realized through non-singular fast terminal SMC. Subsequently, the online distributed collaborative optimization is performed on the sliding mode reaching law, which is designed using the action-critic reinforcement learning method. Consequently, the energy optimality and adaptability of the 3D guidance laws are ensured. Moreover, external disturbances caused by the variable maneuvering target are estimated by non-homogeneous disturbance observers and compensated within the guidance strategy. Finally, the validity, superiority and robustness of the proposed guidance strategy are verified through convergence analysis and numerical simulation.