Carolin Frueh, Purdue University; Bryan Little, Purdue University; John McGraw, J. T. McGraw and Associates
Keywords: Optical Sensors, Sensor Tasking, Multi-Target Tracking
Abstract:
Optical sensors are widely used in the detection, tracking and characterization of space objects. Because they do not require active illumination, they are relatively inexpensive. In recent years, several new optical sensor networks have been built or are in the process of being furnished not only of government organizations, but also an increasing number of commercial users interested in Space Situational Awareness.
Based on the hardware and software that is used in an optical system, their capabilities differ. Capabilities that are of interest are: Detection rate and limit, precision and accuracy of the extracted measurements and the amount of visible sky, respectively the number of detectable objects that can be covered in a given amount of time.
While some capabilities, such as mount slew rate are already specified by the manufacturer, or easy to determine, such as the field of view, others are far less accessible, because they are determined by the combination of factors that are not all independent of each other. Those factors include the telescope hardware, the telescope software, the observed objects and their characteristics, the observation scenario and also the processing scenario (real time, post-processing during the day etc).
At the same time, there has been a thrust in developing methods that seek optimality that depend on those capabilities as input values: Examples are the design of a new sensor network, which includes the number and placement of the sensors with hardware and software specifications, and the use of an existing sensor network with optimal sensor tasking. Even when there is no influence on the tasking, all multi-target tracking methods rely in their model input on values for likelihood and probability of detection, which are, of course, dependent on the sensor capabilities. As a further complication, the object propreties are not known for the vast majority of the objects that are of interest. The publicly available catalog does not provide sufficient insights.
The thrust of the paper is hence a twofold. For one, the limits of knowledge from publicly available sources on the light reflected from the objects is shown and their uncertainties are quantified.
This first step is followed by a rigorous optical sensor model that is introduced. The model is physics-based and introduces in the first step the influence of the object characteristics and the telescope hardware in combination with the physical observation scenario on the capabilities of detection, probability of detection and measurement precision and uncertainty. With the use of Fisher information gain, theoretical limits can be established.
In the subsequent step the influence of the software and the combination of the processing and observation scenario is shown. In the last step, examples are shown. While the examples have to make decisions about specific systems to use, the derivation is explicitly chosen to be non-specific, such that it applies to all optical sensor systems.
Explicit examples are given in the context of the design of a sensor system allowing for complete geosynchronous object coverage and tracking with a simple Probability Hypothesis Filter. For the examples a brute force approach is compared with the advantages of better physics-based modeling even in the absence of exact knowledge on object properties.
Date of Conference: September 17-20, 2019
Track: Optical Systems & Instrumentation