Adding Highly Efficient Scheduling to NASA’s EUROPA Autonomous Planning Architecture
EUROPA is NASA’s premier autonomous onboard spacecraft planning and replanning system. It has been used for an impressive array of missions such as the Mars Exploration Rover (MER), Deep Space 1 (DS1), and Earth Observing System (EOS). Stottler Henke is working with NASA to tightly integrate high speed, high quality Aurora™ scheduling algorithms with EUROPA’s autonomous planning systems architecture. A combined EUROPA-Aurora system would generate more efficient plans and more optimal scheduling. It would enable EUROPA to better handle scheduling problems in its current domains, as well as generate better plans in new domains that have significant scheduling and resource allocation issues. Better schedules will enable more science to be performed with the same time and resources.
There may be no more important mission for the US military than protection from ballistic missile attack. Thus, for any configuration of sensors, it is extremely important to make the most of the collected sensor data. This project uses artificial intelligence techniques to implement human-quality reasoning on object features extracted from sensor data. The ultimate goal is to develop a data fusion system that uses disparate sensor data more effectively to correlate and assess the lethality of ballistic missile defense targets more accurately. By performing human-quality reasoning, the proposed system can perform superiorly to other systems utilizing the same sensor data. This effort includes collecting actual X-band radar and infrared sensor data for specific moving/spinning targets, developing the correlation, fusion, and classification algorithms, and testing the system’s performance with the real data.
Automated planners use symbolic reasoning techniques and models of the planning domain to generate executable plans comprised of scheduled tasks that achieve specified goals in specified contexts. NASA’s Action Notation Modeling Language (ANML) is a relatively new language for specifying planning domain models. For NASA, we are developing a highly visual integrated development environment (IDE) for reviewing, constructing, understanding, testing, and debugging planning domain models expressed in ANML more quickly and effectively. The IDE is implemented using the Eclipse framework. It includes a syntax-aware text editor, validations that detects problems automatically, and built-in queries that streamline model analysis and debugging. The IDE also includes graphical visualizations that summarize action definitions and relationships among actions and variables to help users see important patterns and relationships, so they can understand and debug ANML models more quickly and effectively.
Fault tolerant mission operations require integration of the system diagnosis, planning, and execution functions. For example, when the diagnosis function identifies a failure or degradation in a system component, the planning function must determine whether the available resources can support current plans and, if they cannot, how the plans should be revised. Typically, the diagnosis and planning functions use different models of the system, so integration usually requires translation between models. For NASA, we are developing Intelliface, a suite of knowledge-based tools that will streamline the integration of automated planning, diagnosis, and execution systems within robots, spacecraft, and other autonomous systems. Intelliface will help connect islands of automation provided by diagnosis, planning, and execution systems to ensure that the full benefits of these intelligent systems can be realized.