Model-Based Adaptation for Performance Assured Control of a High-Speed Supercavitating Torpedo

Supercavitation is a state in which a gas cavity substantially encompasses a body immersed in a liquid. The consequence for underwater bodies is that drag over the non-wetted surface is drastically reduced, and this can be exploited to develop rapid reaction defensive weapons for submarines. In addition to propulsion and active flow control requirements to initiate and maintain the supercavity, guidance and control systems are needed to provide the agility for close aboard engagements. An adaptive control design is proposed to assure performance for the inherent nonlinear dynamics of the free surface flow, the interactions between flow, cavity, and afterbody, and the diverse launch conditions and terminal states demanded by the problem. A model-based framework is chosen in which neural network characterizations of the nonlinear dynamics are used in an online control design. Parameter identification and provably stable update laws are then added to adapt the models used for control decision to account for plant uncertainties, and lastly, a composite error control approach is incorporated to adaptively cancel unwanted nonlinearities in the responses. Phase I uses a medium fidelity simulation for controller design and a higher fidelity model for evaluation to further demonstrate the benefits of our adaptive control architecture.