Naomi Owens Fahrner, Ball Aerospace
Keywords: Space Domain Awareness, Resource Allocation, Sensor Tasking, Heterogeneous Sensors
Abstract:
Space Domain Awareness Resource Allocation (SDARA) has gained much interest over the last few years. Given the rapid growth of resident space objects (RSO), this is no surprise. With over 34,000 RSO greater than 10 cm and only a small fraction of that number of observers, allocating observer resources is imperative.
The resources in this SDARA problem are the space domain awareness (SDA) sensors observing targets. The goal of SDA sensor tasking is to maximize the total number of targets observed while also minimizing resource costs. This is done by constructing an objective function and optimizing this against various constraints.
Current objective function formulations of this SDA problem do not address the different costs associated with tasking heterogeneous sensors. Given the variation in possible observers, it is crucial to also consider the tasking cost of each observer. Depending on the observers, it may be more cost effective to task certain sensors more than others given their cost of use.
This paper presents a new objective function for the problem of SDARA as well as a novel algorithm to maximize this new objective function. This SDARA problem aims to maximize the total number of targets seen while minimizing heterogeneous resource costs. For this purpose, targets primarily consist of objects in the GEO-belt, while observers consist of GEO, LEO, and ground-based sensors.
Additionally, this objective function considers real-life constraints such as only considering targets within the observers’ field of view, outside of the earth’s umbra, not blocked by earth, and with high enough SNR. It also has constraints based on slew time and available memory. It is capable of handling multiple heterogeneous observers at once. Each observer has a unique cost associated with its slew time and time to store images.
In order to optimally allocate resources for SDA, a schedule is created for observers over some time interval. This schedule is comprised of tasks for each observer. A task consists of slew and/or collect commands. Each task has an associated score. An optimal schedule is one in which task scores are maximized. This schedule is found by optimizing the objective function.
A novel algorithm will be presented to maximize this proposed objective function. This algorithm is called the ‘block greedy’ algorithm. The block greedy algorithm is a hybrid of the weapon-target-assignment (WTA) and greedy algorithms. Unlike the traditional, local greedy algorithm, it produces an approximate regionally optimal solution. Unlike the WTA, it runs in a tractable amount of time. The block greedy algorithm borrows speed from the greedy algorithm and optimality from the WTA.
The block greedy algorithm is compared to other common algorithms used in solving the SDA sensor tasking problem. These algorithms include the local greedy algorithm, genetic algorithm, weapon-target-assignment algorithm, and a random search algorithm. These algorithms were compared by running all of them on realistic data sets. The block greedy algorithm outperforms these algorithms in maximizing the proposed objective function.
The block greedy algorithm produces a schedule with a much higher schedule score than all other algorithms compared. Additionally, it consistently creates a schedule in which the highest number of targets are seen out of all algorithms. It is not the fastest algorithm, but it is also not the slowest. What is sacrificed in run time using the block greedy algorithm is gained in its higher schedule score.
In summary, this work not only introduces a new, more realistic objective function formulation for the SDA resource allocation problem, but also creates a better algorithm for tasking resources. This new SDA resource allocator both optimizes the total number of targets to be visited and successfully minimizes resource costs.
Date of Conference: September 14-17, 2021
Track: Dynamic Tasking