The research detailed in this thesis explores the deployment of steerable sensors as an efficient means of improving the capability, capacity and timeliness of existing space surveillance systems, to provide superior levels of Space Situational Awareness (SSA) through enhanced sensor management. These improvements are necessary as the world's increasing reliance on spacefaring has brought about the accumulation of an enormous quantity of man-made objects orbiting the Earth. For almost 60 years the number of objects has grown, causing a commensurate increase in the likelihood of destructive collisions involving manned missions and important space assets. To predict and ideally prevent collisions, a number of agencies endeavour to track as many Resident Space Objects (RSOs) as possible. Recent events involving unprecedented surges in the number of RSOs have made it clear that the ability to continue to operate safely in Earth orbit will require enhancements to existing levels of SSA. The act of maintaining SSA is reliant on many sources of information, of which a primary source is the direct observation of RSOs by space surveillance sensors. These observations are utilised to compile and maintain a catalogue of RSO's orbital state estimates that is analysed to determine the likelihood of collision. Surveillance of this environment is a challenging task that currently has a large dependence on legacy systems and techniques that can benefit from modernisation via the introduction of contemporary technologies and methodologies. Due to potential benefits such as low cost, high accuracy, scalability, flexibility and automation, the large scale deployment of steerable sensors is proposed as a means of improving existing catalogue maintenance systems.
Researching a judicious means of deploying and exploiting steerable sensors to improve the capability, capacity and timeliness of space surveillance networks requires consideration of the management of sensors at both a network and an individual level.
The exploration begins at the network level with an analysis of existing practices for maintaining RSO catalogues to understand how catalogue accuracy is affected when steerable sensors are employed. Through numerical simulation, the effectiveness of the current state of the art in steerable sensors, a class of electro-optical sensor, is contrasted with traditional radar surveillance. The findings indicate that if the current state of the art in steerable sensors were to be widely deployed as the primary contributing sensors, catalogue accuracy would increase significantly. The findings also show that greater catalogue accuracy could be expected if effects caused by passive optical sensing to observability of RSO range and sensor availability can be minimised.
Methods for improving observability and availability when using networks of optical sensors are investigated next. Measurement level sensor fusion and efficient analysis of the network's visibility of the RSO catalogue are considered. The investigation's results show that measurement level sensor fusion is capable of reducing catalogue error caused by weak observability of range. However, the effectiveness of the result is highly dependent on the distribution of sensors and ensuring multiple sensors are only tasked to observe a single object when it benefits the catalogue as a whole. Parallel General Purpose computing on a Graphical Processing Unit (GPGPU) is employed to achieve efficient, full-scale simulation and visibility prediction of alternative network configurations. The findings indicate that using a high ratio of optical to radar sensors can achieve high levels of system availability when monitoring a realistic distribution of RSOs. The practicality of the network is further enhanced via visibility prediction by enabling the sensor manager to anticipate small lapses in coverage and schedule observations accordingly. These techniques may be used to implement a surveillance network using steerable electro-optical sensors that overcome the identified constraints to observability and availability.
The final investigation aims to improve upon system capacity and timeliness via enhanced management and control of individual steerable sensors using real-time, GPU-augmented decision making at the sensor. A novel method called dynamic steering is proposed to exploit this architecture and enable a sensor to autonomously switch between tracking and searching whilst reacquiring catalogued objects. Successful, automatic reacquisition was demonstrated in an experimental field trial of the system using targeting data as much as six months old, far surpassing the limits of existing surveillance systems. This result demonstrates an ability to achieve greater levels of system capacity as certain RSOs may be observed less regularly. The autonomy achieved increases the timeliness of the system, as dynamic steering permits the replacement of tasks currently conducted by human operators and reduces the level of tracking refinement necessary before information may be shared between sensors. Such capabilities have the potential to improve existing methods and inspire new techniques for obtaining SSA.
The proposed techniques for managing sensors at the network and individual sensor level may be used to improve the capability, capacity and timeliness of existing space surveillance systems to achieve improved SSA.