MIDAS

The Satellite Control Network (SCN) 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 SCN 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 another 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 downlink scheduling requests, as well as applications to planning (what-if scenarios) and training.

Satellite communications scheduling is a laborious process requiring many schedulers—each with years of training and experience—to meet the current needs of the SCN. 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 SCN 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, equipment expansions, etc. With only a few minutes of processing, MIDAS is able to deconflict a high percentage of tasks in any given day, and this has obvious advantages for current SCN 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 for the strategic exploration of the impacts of various changes at a very high level of detail.

Two MIDAS systems are currently deployed to 22SOPS at Schriever Air Force Base in Colorado Springs, CO 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 previously required). 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 SCN, whose capabilities and requests continually increase, MIDAS will ease the labor-intensive process of scheduling satellite contacts, maintenance, and other related tasks. It will also provide a means for a highly detailed analyses of the possible impact of various changes in the SCN landscape (an unexpected outage or a planned reduction or expansion of capacity), without requiring humans to deconflict the schedule for each possible contingency.

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.

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 SCN 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 SCN.

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.

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.

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, namely MARS, with the goal of developing a distributed, intelligent, and cooperative SCN 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.

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 SCN, and more specifically, individual SOCs.

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.