**ROB MILLER - RESEARCH INTERESTS**

**Academic Background**

I have a BA in Mathematics, an MSc in Computer Science and a PhD in Artificial Intelligence. The title of my doctoral thesis, completed in 1995 at Imperial College, was "Formal Reasoning about Actions and Narratives" and was in the general area of applied logic. Before taking my MSc and PhD I taught mathematics at secondary school and sixth form level for four years. After my PhD I worked as a research associate at Imperial College and then at Queen Mary and Westfield College, before moving to my present post at DIS in 1998.

**Current Research**

My current research is in Artificial Intelligence, and more particularly in applied logic, logic programming, knowledge representation and modelling "common sense" reasoning. I am a member of the Applied Logic Group at UCL, which also includes researchers from the Department of Computer Science. My recent work has concentrated in developing formal languages, theories and computational methods for reasoning about actions, and exploring their potential for application in such areas as automated planning, diagnosis and software requirements engineering. In particular, I am currently working with collaborators at Imperial College, the University of Cyprus and Harvard to investigate how recent work on argumentation may be adapted so as to provide a conceptual and computational framework for automated planning and reasoning about actions.

The two logic-based formalisms that I have been most closely involved in developing and applying in the ways described above are the Event Calculus and Modular-E. The **Event Calculus** is a classical logic formalism for describing and reasoning about the effects of actions. I have worked, for example, in extending this framework so that the AI techniques it encapsulates can be brought to bear on domains involving continuous change. The resulting extended Event Calculus provides an automatic mechanism for reasoning about transitions between the sets of simultaneous differential equations traditionally used to model such dynamic systems, and in particular gives a formal calculus for reasoning about discontinuities of parameters at such transition points. **Modular-E** (along with its predecessor the **Language E**) is an example of an "action description language", with its own tailor-made syntax and model-theoretic semantics. We have used this language as a vehicle to undertake in-depth studies of some of the main current open problems in the field, such as the "frame", "ramification" and "qualification" problems. Our current focus is on the qualification problem, and in particular how to specify how an intelligent agent may cope with or "recover from" observations about the environment that apparently conflict with the agent's own internal model of how the environment should behave.