Autonomous Neural Control of Space Platforms

The autonomous control of flexible space structures will require a distributed computational architecture that provides the ability to perform system identification and adaptive control after orbital deployment. Neural network based supervised learning algorithms and hardware provide a beneficial tool for reducing the need for a priori knowledge of the structures behavior through rapid system identification and health monitoring on-line. Furthermore, the combined use of neural networks and optimal control presents a stochastic nonlinear learning control architecture that provides the ability to rapidly adapt on-line to uncertainty in the structure's behavior. Under an existing program from McDonnell Douglas Aerospace, NeuroDyne is applying internally developed decentralized learning optimal control tools to active vibrational damping of "smart" composite aircraft structures. In Phase I, our team will leverage this research to further develop and apply real-time system identification and learning optimal control for control/structural interaction (CSI) of the Phillips Laboratory Planar Articulating Controls Experiment (PACE) test bed. Final goal of this effort will be integration of controls and demonstration of vibration damping and precision pointing for the PACE structure.