This paper addresses the problem of singularity avoidance in a four-control-moment-gyroscope cluster as used for the attitude control of a satellite. A global search algorithm is developed that adjusts the null motion added upon the singularity robust inverse steering law. Its principal characteristic is that it uses global information gathered from the whole maneuver, compared to most conventional techniques that consider only some local information, near the current gimbal configuration for the optimization. The method is implemented using a ternary tree structure, and a heuristic algorithm optimizes a cost function that depends on the manipulability index, the time spent in the vicinity of singularity, the quaternion error, and the gimbal rates. Specific measures are taken to decrease the execution time of the algorithm, even for long maneuvers, without the need of a visit histogram. In addition, the tuning of only two variables can drastically change the execution time and the computational resources needed. The numerical simulation used to evaluate the algorithm indicates that the elliptic singularity can sufficiently be avoided and the algorithm drives the system fast away from the singularity with an overall performance improvement.

本文讨论了用于卫星姿态控制的四控制力矩陀螺仪群中避免奇点的问题。开发了一种全局搜索算法，该算法可调整在奇异性鲁棒逆转向定律上添加的零运动。它的主要特征是，与大多数只考虑一些局部信息的常规技术相比，它使用了从整个操纵过程中收集的全局信息，而该优化技术接近于当前的万向架配置。该方法使用三叉树结构实现，并且启发式算法优化了成本函数，该函数取决于可操作性指标，在奇点附近花费的时间，四元数误差和万向节速率。采取了特殊措施以减少算法的执行时间，即使是长时间的操作，无需访问直方图。此外，仅调整两个变量可以大大改变执行时间和所需的计算资源。用于评估该算法的数值模拟表明，可以充分避免椭圆奇异性，并且该算法使系统快速远离奇异性，从而改善了整体性能。