The rapid growth of man-made objects in Earth’s orbit has created a need for planning and prioritizing sensor measurements of these objects in order to keep track of their motion in orbit. In this paper, we adapt a concept from the field of system identification, namely application-oriented input design, to the sensor-planning problem. Previous methods have predominantly focused on the problem of planning the measurements of different objects such that they are as accurate as possible (in a broad sense) from a given set of possible observations. Here we view the problem differently; the objective is to find the least costly number of measurement such that the estimated orbital parameters are good enough for their intended application. To this end, we introduce the concept of an application set. The application set contains all orbital parameters that are considered of sufficient accuracy for the intended use. The objective of the optimization is hence not to find the most accurate set of parameters but find the cheapest combinations of observations such that the estimated parameters are guaranteed to be within the application set with a high probability. We show how to formulate the planning problem as a convex optimization problem that can be solved efficiently with modern algorithms, even for large-scale problems. We then demonstrate the feasibility of the method in a simulation example. Finally, the paper discusses some interesting topics for future research.

地球轨道上人造物体的快速增长需要对这些物体的传感器测量进行规划和优先排序，以便跟踪它们在轨道上的运动。在本文中，我们将系统识别领域的概念（即面向应用的输入设计）应用于传感器规划问题。以前的方法主要关注规划不同对象的测量的问题，以便根据给定的一组可能的观察结果尽可能准确（广义上）。在这里我们以不同的方式看待这个问题；目标是找到成本最低的测量次数，以使估计的轨道参数足以满足其预期应用的需要。为此，我们引入应用程序集的概念。该应用程序集包含所有被认为对于预期用途具有足够精度的轨道参数。因此，优化的目标不是找到最准确的参数集，而是找到最便宜的观测组合，以便保证估计的参数以高概率位于应用程序集中。我们展示了如何将规划问题表述为凸优化问题，即使对于大规模问题也可以使用现代算法有效解决。然后我们在仿真示例中证明了该方法的可行性。最后，本文讨论了未来研究的一些有趣的主题。