A Machine Learning Approach to Enable Mission Planning of Time-Optimal Attitude Maneuvers

Time-optimal spacecraft rotations have been developed and implemented on orbiting spacecraft, highlighting opportunities for improving slew performance. Double-digit reductions in the time required to slew from one attitude to another have been demonstrated. However, the ability to perform mission planning to make use of minimum time slewing maneuvers is largely precluded by the need to compute a numerical solution to find a single minimum time maneuver control trajectory. Machine...
More Description

Time-optimal spacecraft rotations have been developed and implemented on orbiting spacecraft, highlighting opportunities for improving slew performance. Double-digit reductions in the time required to slew from one attitude to another have been demonstrated. However, the ability to perform mission planning to make use of minimum time slewing maneuvers is largely precluded by the need to compute a numerical solution to find a single minimum time maneuver control trajectory. Machine learning approaches can eliminate the need to generate problem solutions by approximating time-optimal maneuver times with sufficient accuracy for planning using only the initial and final attitude requirements. The advantages of time-optimal spacecraft maneuvers, a planning construct for evaluating legacy and machine learning maneuver time generators, and the machine learning processes that enable this approach are outlined. Compared to legacy planning techniques, time-optimal slew approximations yield target collection increases of 3% to 24% for an example planning framework.