Computer Engineering Commons^™

Parameter Analysis Of An Adaptive, Fault-Tolerant Attitude Control System Using Lazy Learning, Matthew Russell

Senior Honors Theses

Several key requirements would be met in an ideal fault-tolerant, adaptive spacecraft attitude controller, all centered around increasing tolerance to actuator non-idealities and other unknown quantities. This study seeks to better understand the application of lazy local learning to attitude control by characterizing the effect of bandwidth and the number of training points on the system’s performance. Using NASA’s 42 simulation framework, the experiment determined that in nominal operating scenarios, the actuator input/output relationship is linear. Once enough information is available to capture this linearity, additional training data and differing bandwidths did not significantly affect the system’s performance.