- MIDAS – Managed Intelligent Deconfliction and Scheduling for the Air Force Satellite Control Network
- SSN Scheduling – Improved Space Surveillance Network (SSN) Scheduling
- MALINA – Automated Performance Assessment for Training Satellite Planners Based on Learned Metrics
- TRACER – Intelligent Terrestrial EMI Emitter Locator for AFSCN Ground Stations based on AI Techniques
- RAPTOR – RFI Detection and Prediction Tool
- PASAP – Optimization of Communication Networks with Geodesic Dome Phased Array Antennas using Artificial Intelligence Techniques
Managed Intelligent Deconfliction and Scheduling for the Air Force Satellite Control Network (AFSCN)
Scheduling and Deconflicting AFSCN Communications
The Air Force Satellite Control Network (AFSCN) coordinates hundreds of satellite communication requests from various users every day. MIDAS (Managed Intelligent Deconfliction and Scheduling) is a tool for rapidly scheduling and deconflicting AFSCN satellite communication requests. In the past, these needs were exclusively met by teams of highly trained and experienced schedulers manually checking every schedule request received. Approximately half of all requests require adjustment to remove conflicts. MIDAS now automates much of this, allowing schedulers to apply their expertise where it is really needed. It accomplishes this with a two-stage process that first shuffles tasks within their defined constraints before carefully applying a user-definable set of business rules that allow certain constraints to be relaxed when necessary. The system provides a familiar, user-friendly interface modeled on legacy Electronic Schedule Dissemination (ESD) systems to facilitate comparison and to allow users to switch from one interface to the other with relative ease. It runs on inexpensive consumer hardware and communicates with legacy systems via a well-defined plain-text file format: raw scheduling requests are imported to MIDAS, and scheduling results can be exported back to legacy tools. This tool is useful for rapidly deconflicting real-world scheduling requests but also has possible applications to planning (what-if scenarios) and training.
Leveraging Human Knowledge and Expertise to Reduce the Burden on Human Schedulers: An Artificial Intelligence Solution
Satellite communications scheduling is a laborious process requiring many schedulers—each with years of training and experience—to meet the current needs of the AFSCN. An artificial intelligence (AI) tool for automatically scheduling satellite contacts, designed to incorporate the experience, insights, and expertise, and mimic the thought processes, of human schedulers–including incorporating a unique precedence-based capability to “bend the rules” the way veteran schedulers do–will take much of the burden off of these human schedulers, allowing scheduling to scale along with AFSCN capabilities. Additionally, an automated scheduling tool allows planning at a level not previously possible. A viable schedule can be assembled in a matter of minutes in order to assess the impact of possible outages, events, expansion of equipment, etc. With only a few minutes of processing, MIDAS is able to deconflict a high percentage of tasks in any given day. This has obvious advantages for current AFSCN operations. MIDAS can eliminate much of the repetitive work involved in scheduling and allow schedulers to focus on difficult problems.
Future versions of MIDAS could also improve response time in the event of a mission change or emergency. Additionally, this tool provides an inexpensive way to explore hypothetical scenarios, allowing strategic exploration of the impacts of various changes at a very high level of detail.
Rapid Scheduling and Deconfliction: The MIDAS Touch
Two MIDAS systems are currently deployed to 22SOPS at Schriever Air Force Base in Colorado Springs for testing and verification. Users can export data from their legacy scheduling tool (ESD 2.7x) to MIDAS, where they can generate a schedule solution with most conflicts resolved in about 5 minutes. Air Force results demonstrate that this intermediate result can be fully deconflicted by working with Aurora’s suggestions in about 30 minutes by a single scheduler (compared to approximately 9 person-hours required previously). Through further refinement, we expect the number of initial conflicts resolved to increase and the manual deconfliction required to be reduced even further. We expect MIDAS to be useful in current scheduling operations and to be one component of 22SOPS’s scheduling technology in the future. Because MIDAS is not tightly integrated with any one system, it can be customized to interoperate with any system, including the current legacy ESD system and the planned ESD 3.0.
With an increasingly demanding AFSCN, with both capabilities and requests increasing, MIDAS will ease the labor-intensive process of scheduling satellite contacts, maintenance, and other related tasks. It will also provide a means for highly detailed analyses of the possible impact of various changes in the AFSCN landscape (an unexpected outage or a planned reduction or expansion of capacity), without requiring humans to deconflict the schedule for each possible contingency.
Automating SOC Contact Scheduling
Stottler Henke has built on the success of the core MIDAS system to develop a new system for the Air Force that improves the process of scheduling communication “contacts” between Satellite Operations Centers (SOCs) and Remote Tracking Stations (RTSs) on the ground, and satellites in orbit.
This weekly planning process, known as Program Action Plan (PAP) generation, begins at individual SOCs where mission operators must manually select a handful of suitable supports from a list of hundreds of available time segments per day, without regard for the scheduling requirements of other missions. Once the desired supports are selected for a particular mission, the PAP is delivered to the 22nd Space Operations Squadron (22 SOPS) for centralized scheduling and global deconfliction across all SOCs. This results in frequent conflicts and the need to make schedule changes, often requiring manual adjustments and approval by the originating SOC.
The MIDAS 3.5 effort sought to employ intelligent automation to automatically generate a high-quality PAP, thereby reducing the required SOC manpower and setting the stage for the eventual transition to “lights out” operation at the SOCs.
Generating and Deconflicting Program Action Plans (PAPs) Using Artificial Intelligence (AI)
MIDAS 3.5 automates the PAP generation process by analyzing the requirements for one or more missions supported by a SOC and then automatically generating a PAP of supports that meet the requirements for the necessary satellite contacts.
Stottler Henke designed MIDAS 3.5 in close collaboration with mission operators at a dedicated SOC (the RDT&E Support Complex (RSC)) at Kirtland AFB in Albuquerque, NM. This provided us insight into the complex requirements, constraints, and scheduling rules of thumb intrinsic to AFSCN scheduling, as well as common conflicts and how they are resolved. The resulting system incorporates heuristic algorithms based on the scheduling behavior of real human operators in order to rapidly generate a high-quality PAP for individual missions.
Testing and evaluation at the RSC showed that with only a few minutes of processing, MIDAS 3.5 is able to automatically select contacts that meet 100% of mission requirements. In fact, the system will often generate a more complete schedule (more viable contacts per day) than those created by human operators. MIDAS 3.5 is currently being considered for transition to operational use at the RSC. With relatively simple and straightforward customizations, MIDAS 3.5 could be used by other SOCs within the AFSCN.
The system provides a modern, user-friendly interface to clearly and intuitively specify mission requirements and review the generated PAP and any potential problems. MIDAS 3.5 runs on inexpensive consumer hardware and communicates the selected contacts and visibility data via the familiar PAP and SAT file formats, as well as via DEFT files—a well-defined plain-text file format that is also read and written by ESD 2.7 and other related tools. Visibility data for a given planning period is imported to MIDAS 3.5 from Standard Orbit Events Table (SORBET) files that contain contact-specific data identifying the visibilities during which a satellite can be contacted.
Facilitating “Lights Out” Operations
MIDAS 3.5 can serve as a stepping stone to help transition to fully automated “lights out” SOC operations with MMSOC 2.1. As capabilities, capacity, and request volume all increase, automation in this area will soon become a necessity.
By interfacing with MMSOC 2.1 via a GMSEC (Goddard Mission Services Evolution Center) compliant adapter, MIDAS 3.5 could automatically send and receive the messages necessary to generate local SOC schedules and communicate with central scheduling at the 22nd Space Operations (22 SOPS).
This would allow MIDAS 3.5 to automatically ingest the SORBET file that identifies the available visibilities for a mission, automatically generate the corresponding PAP, and publish a message containing the selected supports to 22 SOPS—which would not only reduce the manpower required to generate a local SOC schedule, but simultaneously improve the chances of global optimization without the manual change/review/accept cycles between SOCs and 22 SOPS.
A Fully Integrated Solution
The MIDAS 3.5 system is already able to communicate the generated PAP to the core MIDAS system (an intelligent scheduling agent for 22 SOPS) via the familiar PAP, SAT, and DEFT file formats.
Furthermore, development of MIDAS 3.5 is directly applicable to a related effort, MARS, with the goal of developing a distributed, intelligent, cooperative AFSCN mission planning and scheduling system to improve the overall scheduling process and provide greater sharing and availability of information and knowledge.
The MIDAS 3.5 technology allows for a cooperative solution to be negotiated among distributed agents, specifically MIDAS 3.5 agents at individual SOCs and the core MIDAS system at 22 SOPS (described above). MIDAS 3.5 could selectively transmit additional scheduling metadata and constraints from the SOC for use during centralized scheduling at 22 SOPS. This additional information would allow the core MIDAS system at 22 SOPS to make more informed suggestions with a higher likelihood of being approved by the SOC, thereby reducing the amount of back-and-forth negotiation required between 22 SOPS and individual SOCs.
A Wealth of Use Cases
MIDAS 3.5 is a powerful tool that can be applied to a variety of SOC scheduling situations.
For instance, MIDAS 3.5 can be used as an excellent training tool, allowing a trainee to work without an actual ESD terminal, and it provides additional benefit by suggesting possible solutions to inexperienced schedulers who may not be familiar with all possible moves.
Additionally, MIDAS 3.5 can be used for “what-if” studies to explore, for example, the consequences of a particular outage or failure. The manpower to ascertain the impact of such changes without MIDAS 3.5 is prohibitive.
Finally, MIDAS 3.5 can serve as a stepping stone to help transition to fully automated “lights out” SOC operations with MMSOC 2.1. As capabilities, capacity, and request volume increase, automation in this area will become a necessity.
The most direct benefit of this effort is, of course, providing intelligent automation for the AFSCN and, more specifically, individual SOCs.
Beyond the Air Force: Additional Government, and Commercial Applications
This technology is already proving useful to the Air Force, and there exists strong commercialization potential for MIDAS 3.5 with NASA and private sector satellite operators. Additionally, there are many similarities between communication scheduling and sensor scheduling. This effort resulted in additional scheduling algorithms and heuristics that have been incorporated into our existing scheduling tool Aurora. Similarly, the scheduling algorithms developed for this product could be applicable to additional parties within the Air Force requiring this type of capability.
Improved Space Surveillance Network (SSN) Scheduling
There are close to 20,000 cataloged manmade objects in space, the large majority of which are not active, functioning satellites. These are tracked by phased array and mechanical radars and ground and space-based optical telescopes, collectively known as the Space Surveillance Network (SSN). A better SSN schedule of observations could, using exactly the same legacy sensor resources, improve space catalog accuracy through more complementary tracking, provide better responsiveness to real-time changes, better track small debris in low earth orbit (LEO) through efficient use of applicable sensors, efficiently track deep space (DS) frequent revisit objects, handle increased numbers of objects and new types of sensors, and take advantage of future improved communication and control to globally optimize the SSN schedule. We have developed a scheduling algorithm that takes as input the space catalog and the associated covariance matrices and produces a globally optimized schedule for each sensor site as to what objects to observe and when. This algorithm is able to schedule more observations with the same sensor resources and have those observations be more complementary, in terms of the precision with which each orbit metric is known, to produce a satellite observation schedule that, when executed, minimizes the covariances across the entire space object catalog. If used operationally, the results would be significantly increased accuracy of the space catalog with fewer lost objects with the same set of sensor resources. This approach inherently can also trade-off fewer high priority tasks against more lower-priority tasks, when there is benefit in doing so.
Automated Performance Assessment for Training Satellite Planners Based on Learned Metrics
When planners at Satellite Operations Centers (SOCs) develop schedules for requested satellite contacts, their decision-making must account for a range of planning constraints, specialized for different satellites. In order to train SOC planners in simulation-based scenarios, performance assessment measures are needed to evaluate their planning decisions in context and provide feedback. These assessment measures can be complex and time-intensive to develop, especially when there are many different satellites and SOCs, with different sets of mission-related constraints.
Stottler Henke is developing a set of authoring tools called Modular Annotated Learning for Instructional Authoring (MALINA) that provides a novel collaborative approach to constructing assessment knowledge, by combining machine learning with instructor-provided annotations. Input data include sample solutions collected from real-world operations, logged training events, or instructor-created examples. MALINA provides an annotation utility to supplement learned assessment measures with additional information from a human instructor or expert. This is particularly important for cases where there are soft rules that a human expert may follow, and that the system must learn. The system then induces performance assessment knowledge from the annotated sample data. MALINA is conceived as a domain-independent approach to building automated assessment for simulation-based training. However, the domain of SOC planning is the initial development application for the MALINA approach with an exemplar training prototype. The MALINA authoring methodology is being applied first with a specific SOC using accompanying knowledge of the specific constraints for the satellites managed by that SOC. A second training application with a different SOC will present an opportunity to explore generalizability to the planning decisions involved with different satellites and different constraints.
The MALINA learning mechanism also presents an additional potential benefit beyond the training context. By learning the constraints used in decision-making for a given domain such as SOC satellite planning, these constraints can be employed not only for automated performance assessment in a training setting, but also for suggesting possible solutions to challenging cases in an operational setting.
Intelligent Terrestrial EMI Emitter Locator for AFSCN Ground Stations based on AI Techniques
TRACER (Terrestrial RFI-locating Automation with CasE based Reasoning) is an AI-based, automated, electro magnetic interference (EMI) emitter localization and identification system. TRACER investigates suspected terrestrial sources of EMI, presents Air Force personnel with an intuitive description and classification of each EMI incursion and its current impact on operations, and recommends which steps should be taken to mitigate the interference.
To detect EMI, TRACER monitors signals from Remote Tracking Station and Directional Finder antennas around the world, via an interface to ALPS, a system developed by Intelligent Software Solutions (ISS). Upon first confirming or excluding the presence of space-based sources of interference, TRACER—based on details of the signal such as its classification (as determined by neural network technology developed by Data Fusion & Neural Networks (DF&NN), direction, and strength—retrieves and implements one or more investigative methodology cases. TRACER leverages Behavior Transition Networks from Stottler Henke’s SimBionic to rapidly and continuously update the probabilities of various possible sources of EMI. For instance, should one hypothesis be that the EMI originates from an aircraft, a first bearing and estimated distance can be plotted on Google Earth, a real-time flight tracking site (such as FlightAware) accessed, and the flights in the air at the time of the first EMI data point plotted. When the earth-based antenna has made its second sweep, the amount of bearing change confirms or refutes the hypothesis.
TRACER will enhance Space Situational Awareness (SSA), offer increased support for adversarial scenarios (both real and during training), and dramatically shorten EMI response time—and, in so doing, realize significant manpower savings.
RFI Detection and Prediction Tool
RAPTOR is an enhancement of Stottler Henke’s intelligent scheduling tool, Aurora, for improving Air Force satellite scheduling processes and providing greater sharing and availability of information and knowledge among SOCs, automating the entire AFSCN process. RAPTOR provides better-quality schedules, faster scheduling, and the ability to handle larger, more complex sets of requests. Using artificial intelligence methods and techniques, RAPTOR negotiates resolution of conflicts in an automated or semi-automated manner and performs far-future and contingency scheduling/planning as well as automatic abnormal real-time scheduling signal detection and prediction. Its intelligent, intuitive user interfaces enable graphical editing and management of decision processes associated with both satellite constellations and individual satellites by SOCs, Real-Time Schedulers (RTSs), the 22 SOPS, and the 22 SOPS crew commander. RAPTOR complements the successes already achieved using MIDAS to deconflict schedule requests across all SOCs for 22 SOPS.
RAPTOR performs AFSCN scheduling, deconfliction, and negotiation both during the deconfliction period and in real-time; improving and enhancing ABNet to increase its level of automation, greatly decreasing the amount of manpower required, incorporating additional inputs, and providing reasoning over its outputs for intuitive, concise displays to end-users; allows appropriate far-future and contingency planning; provides intelligent, user-friendly interfaces for a diversity of personnel; and integrate with existing required software and prepares to integrate with the Joint Space Operations Center (JSpOC) Mission System (JMS).
The most direct target for the results of this effort is the AFSCN. Additionally NASA has the space communications networks, which can utilize this software (Deep Space Network (DSN), Space Network (SN), and Ground Network (GN)).
Optimization of Communication Networks with Geodesic Dome Phased Array Antennas using Artificial Intelligence Techniques
Stottler Henke’s Phased Array Smart Allocation and Planning (PASAP) tool applies artificial intelligence techniques to two distinct, yet interrelated problems with planning Geodesic Dome Phased Array Antenna (GDPAA) contacts. First, PASAP comes equipped with a smart beam allocation algorithm, which is responsible for the task of allocating beams to transmit/receive modules on the antenna without overloading. Moreover, PASAP features a beam path planning algorithm. This path planning algorithm was originally developed for the U.S. Army to enable UAVs and piloted rotary wing aircraft to more capably avoid threats and obstacles; for PASAP, we modified this algorithm to meet the challenge of better beam path planning — demonstrating the flexibility of Stottler Henke’s algorithms and how we often customize the same basic algorithms or software systems to address a diversity of customers and their unique needs.
Given an allocation of beams to transmit/receive on the antenna, a path across the topography of a geodesic dome antenna surface must be established that preserves the assignments without overloading or violating communication constraints. Although the PASAP tool was developed with the specific constraints of existing GDPAA designs in mind, it is intentionally abstract in order to apply for a broader range of potential constraint satisfaction problems with different kinds of phased arrays. The initial GDPAA application was designed to contribute to the overall feasibility of GDPAA utilization by demonstrating an automated means of optimizing the increased capacity that comes with such a platform, which holds the potential to not only reduce the manpower requirements for the scheduling activity, but also to contribute to greater overall satellite communication capacity.