The broad objective of this CAREER project is to help cement the status of software engineering as a discipline that uses models, methods, and mechanisms based on generalizable scientific knowledge. The research component of this project addresses three important problems in software design. The mediator method is an approach to design in which the behavioral relationships needed to integrate independent behaviors are separately identified and represented as objects external to the objects representing those behaviors. Externalizing relationships as mediators depends in turn on a careful application of both explicit and implicit invocation mechanisms. The first research focus is an attempt to improve scientific foundations for software design, scientifically testing the claim that common methods for designing integrated systems are poorer than the mediator method with respect to ease of evolution of integration requirements. Second is the exploration, through pilot projects, of using the mediator method to aid the design of distributed systems. Third is the investigation of a new approach to helping designers balance multiple objectives as software designs evolve. An environment that reflects how well a design meets each of several objectives by depicting divergences of the design from each of several a models, where each model is defined by the designer as good with respect to one particular objective. The teaching focus is intended to develop engineers who understand software design to be a multi-objective, risky activity, done with incomplete information and thus demanding creativity and reflective, rigorous, risk-reducing methods. To help students understand how to manage the complexity and risk of real software systems, courses focus on realistic development projects, such as a mobile robot system.
