Adaptive Critic Neural Network based Terminal Area Energy Management/Entry Guidance

Reusable Launch Vehicles (RLVs) are the future of the space industry due to their low cost and reliability. The new generation of RLVs have different mission requirements than the space shuttle, which is used for benchmark guidance; therefore, various guidance schemes should be investigated in the interest of further cost reduction and safety. Terminal Area Energy Management (TAEM) is the final phase of entry flight, which is loosely defined to begin at a velocity of 2500 ft/s. Currently the guidance trajectory that exists in this flight phase severely limits the allowable vehicle states for a successful landing. This paper presents a methodology that generates numerous TAEM guidance trajectories, which allow a wide range of vehicle states during landing. The methodology developed to generate these trajectories is a neural network based solution for a finite horizon trajectory optimization problem. Adaptive-Critic based neural networks, a dual neural network formulation, generate the trajectories. The cost function is formulated so as to maintain a gradual glideslope and meet specific terminal constraints. Pointmass time dependent dynamics of a reusable launch vehicle are used. The nonlinear dynamics were reformulated to be dependent on downrange distance since the final time of the trajectory is not always known. Numerical results and interpretations of the associated dynamics are discussed