Abstract
The need to improve the quality of health care has led to a strong demand for clinical protocols and computer systems supporting both their creation and execution. Current approaches in the planning community concentrate on algorithmic improvements, but mostly fail in medical applications. Planning approaches are based on assumptions like deterministic behavior, which do not hold in medical domains. Additionally, they do not cover the problem area of acquisition and verifying complex domain knowledge. In the field of medicine there is a strong movement towards clinical protocols – resembling plans in AI – but computer support during execution of these plans (e.g., controlling, selection of alternatives) is still basic. We need to build complex plans, but also to reason about them in different ways in order to modify plans, to consider the effects of different plans over time, and to monitor execution. We call this range of reasoning tasks plan management. We describe the requirements for these intertwined tasks of plan management so as to respond to the practical demands, to compare approaches in planning and medical informatics to these requirements, and finally, to discuss how our Asgaard project can meet them.