Venue: 94th AMS Annual Meeting 2014
The transition of algorithms and software from a research environment to operations is a complex process requiring a methodical series of steps to ensure success. As new systems are proposed and developed, the rapid integration of new technologies in the form of scientific and data processing algorithms are an integral part of the successful deployment of everything from national ground processing facilities to targeted end-user applications. The National Research Council Committee on NASA-NOAA Transition from Research to Operations recommends improved transitional processes for bridging technology from research to operations. The three key elements of this process are: (a) a well-defined algorithm development process that is shared by the science/engineering teams; (b) a two-tiered software framework comprising development/algorithm engineering and production environments, which share common infrastructure/interface elements; and (c) a common service-based architecture that promotes net-centric operational paradigms.
Effective transition of research algorithms to operations requires a close partnership between the scientists who develop the underlying science and proofs of concept and the software engineers who transition these proofs of concept to reliable, maintainable and validated operational systems meeting latency and software quality requirements. This process is made effective by involvement of the scientists at each stage of the software life-cycle, but in a controlled manner as specified by the integrated “algorithm engineering” process.
The shared data model interface provides common methods for integrating test and a seamless mechanism for transitioning algorithms to operations. In addition, it also promotes algorithm “buy back” from the operational environment to the scientific development environment. “Buy back' enables science algorithm developers to add new capabilities and improvements to the operational software in a flexible and convenient development environment, streamlining transition to operations of the upgrades.
The common data model and interfaces are key design features enabling encapsulation of algorithms as well-defined components. The proposed architectural approach features a programmatically-accessible database of algorithm and data characteristics, which can reduce redundancy and implementation errors across a wide variety of missions and applications, and can be employed out-of-the-box on all stages of ground system life-cycle from design and development to test and operations. Usage of this algorithm oriented database paradigm also enables the creation of powerful systems engineering and auditing tools that can be leveraged across multiple missions. The approach also allows a mission to better adapt to changes and reduces the risk of documentation becoming out of sync.
In this paper we describe an overall framework for developing a research to operations transition plan, provide examples of the successful transition of diverse algorithms for multiple customers, and discuss recommendations for how the both the research community and operational centers could work together to ensure a more efficient transition. We also describe the underlying software architecture that supports this approach. This work provides several illustrative examples , and outlines how these processes could enable future design of ground processing systems for the production of environmental remote sensing products.