An adaptive neuromorphic chip for augmentative control of air breathing jet turbine engines
John C. Gallagher, Kshitij S. Deshpande, Mitch Wolff · 2008
Continuous Time Recurrent Neural Network Evolvable Hardware (CTRNN-EH) has been proposed as an enabling control technology for electromechanical devices. In addition to being able to learn control laws tabula rasa, CTRNNs can learn how to augment existing, trusted, controllers to add new capabilities without breaking existing operation. The ability to augment would be most useful in situations in which significant patching of existing controllers is needed to address contingencies not seen at design time and in which traditional design processes might be too slow to deliver quickly. In this paper, we will discuss the use of CTRNN-EH to augment a standard FADEC controller for an air-breathing jet turbine engine. We will show how we were able to extend the FADEC to properly control thrust under unusual loading conditions that were not considered at design time. Following, we will discuss future applications.