Dynamic Steering for Improved Sensor Autonomy and Catalogue Maintenance

A number of international agencies endeavour to maintain catalogues of the man-made resident space objects (RSOs) currently orbiting the Earth. Such catalogues are primarily created to anticipate and avoid destructive collisions involving important space assets such as manned missions and active satellites. An agency’s ability to achieve this objective is dependent on the accuracy, reliability and timeliness of the information used to update its catalogue. A primary means for gathering this information is by regularly making direct observations of the tens-of-thousands of currently detectable RSOs via networks of space surveillance sensors. But operational constraints sometimes prevent accurate and timely reacquisition of all known RSOs, which can cause them to become lost to the tracking system. Furthermore, when comprehensive acquisition of new objects does not occur, these objects, in addition to the lost RSOs, result in uncorrelated detections when next observed. Due to the rising number of space-missions and the introduction of newer, more capable space-sensors, the number of uncorrelated targets is at an all-time high. The process of differentiating uncorrelated detections caused by once-acquired now-lost RSOs from newly detected RSOs is a difficult and often labour intensive task. Current methods for overcoming this challenge focus on advancements in orbit propagation and object characterisation to improve prediction accuracy and target identification. In this paper, we describe a complementary approach that incorporates increased awareness of error and failed observations into the RSO tracking solution. Our methodology employs a technique called dynamic steering to improve the autonomy and capability of a space surveillance network’s steerable sensors. By co-situating each sensor with a low-cost high-performance computer, the steerable sensor can quickly and intelligently determine where to steer in order to improve its utility. The sensor-system uses a dedicated parallel-processing architecture to enable it to compute a high-fidelity estimate of the target’s prior state error distribution in real-time. Negative information, such as when an RSO is targeted for observation but it is not observed, is incorporated to improve the likelihood of reacquiring the target when attempting to observe the target in future. The sensor is consequently capable of improving its utility by planning each observation using a sensor steering solution that is informed by all prior attempts at observing the target. We describe the practical implementation of a single experimental sensor and offer the results of recent field measurements and simulations. The proposed approach is applied to the task of Initial Orbit Determination (IOD). By first developing and incorporating a constrained admissible region (CAR), the system is capable of reacquiring an RSO months after it was briefly observed and in spite of the apparent lack of tracking information. The system consequently offers a means of enhancing surveillance for Space Situational Awareness (SSA) via increased system capacity, a higher degree of autonomy and the ability to reacquire objects whose dynamics are insufficiently modelled to cue a conventional space surveillance system for observation and tracking.