Brian C. Williams
In this talk I will examine the lessons learned from DSI about model-based programming and modeldirected execution. In the first part I will examine the representational needs of modeling DS1 as a hybrid system. From this I will develop the Reactive Modelbased Programming Language (RMPL). RMPL uses transition systems and co-temporal interactions as a rallying point for unifying representational concepts from a diverse set of research areas, including qualitative modeling, model-based diagnosis, hidden Markov decision processes, synchronous reactive languages and concurrent constraint programming. In the second part I will discuss the development of a series of increasingly more expressive modeldirected executives. Each executive is formulated as a deductive form of an optimal, model-based controller in which models are specified through a combination of hierarchical, probabilistic transition systems and propositional logic. This framework allows us to achieve high levels of autonomy and responsiveness, by drawing inspiration from a diverse set of algorithms taken from model-based diagnosis, hidden Markov processes, search, planning and real-time propositional inference. I will conclude by discussing the role model-directed autonomy is playing within a variety of future NASA applications, from Mars Exploration to the search for Earth-like planets around other stars.