Computational missile guidance framework based on model predictive path integral

Abstract

We address missile guidance problems with practical constraints such as impact angle, impact time, the seeker's field-of-view (FOV) limit, and the missile's maneuvering limit by using model predictive path integral (MPPI), which would be a new computational missile guidance framework with the potential for in-flight control guidance. The MPPI algorithm optimizes the current control command set using the expected cost of the sampled state trajectories generated by perturbing the current control command with random disturbances, based on the given model. In this study, the current guidance command has the form of proportional navigation (PN) guidance with line-of-sight (LOS) range-varying gain, which is optimized by the iterative update law of the MPPI algorithm at every guidance cycle during flight. The proposed guidance approach has the potential to expand the scope of guidance problems because it does not require accurate estimation of time-to-go or the efforts of convexification procedures for engagement kinematics or constraints. This sets it apart from other existing guidance approaches. Additionally, it does not require an optimal solver, unlike the computational convex optimal approaches. And the real-time closed-loop computational guidance command makes the system robust to environmental uncertainty. The proposed computational missile guidance algorithm's effectiveness and feasibility are verified through numerical simulations.