A system that is aware of the context in which a person is operating and can act accordingly is said to display ambient intelligence. Such systems may involve a number of heterogeneous devices communicating with each other via a variety of protocols. Many of these devices have limited memory and processing power. Moreover, there is a high turnover in device architectures as the field expands.

This paper describes a particular software development environment that allows a designer to build an abstract model of an ambient intelligence system and then generate code for the variety of devices making up that system. It provides a detailed case study of an application in which a user riding a bus can be kept informed about the optimal strategy for reaching his destination on the basis of the actual locations of all buses at the moment. The example is carried through the development process, with explanations of how the appropriate model is constructed and transformed. The authors also report performance measures on the final implementation.

I had trouble at first with the large number of unfamiliar (to me) acronyms, but after some acclimatization I was able to follow the arguments without too much difficulty. The paper gives a reasonable overview of the application area, and the description of the case study is quite complete. There are extensive references that supplement the description of the development environment and summarize related work.