As computers decline in cost and interactive computation moves off of the desktop, computation can assume a variety of physical forms. At present, the creation of new computer interfaces is limited to already-available formats, such as smartphones, and even then is constrained to relatively simplistic computer interfaces. Computing in the future will push beyond desktops, laptops, tablets and smartphones into objects that fit into every part of our lives. But for this to happen, design tools are needed that can rapidly adapt computing to different physical forms and task needs. This project will study toolkits for the development of interactive physical objects. The project will increase national competitiveness in the invention and development of new uses of computation, and make possible rich and diverse usability research with physical computing objects in the wild rather than in the lab. Creating physical computing objects will allow a greater proportion of society to use computers in ways that fit their work requirements in fundamentally new ways. The rapid design and creation of such devices will open new markets for consumer computing.
These ends are achieved through three interrelated efforts: (1) A pluggable toolkit for creating sensor/actuator systems that form the computational basis for such objects. This toolkit will be unique in that the components when plugged together not only form an initial prototype of the device but self-reveal their physical and computational characteristics to automatically support the other design tools. A prototype constructed with this toolkit will inherently contain sufficient information to drive its own fabrication. By plugging together the prototype our tools will automatically know how to perform a custom fabrication of the electronics at a size that can be readily deployed. (2) Development of physical form via 3D printing such that the physical shape, sensors and actuators integrate with the software and electronics so as to easily prototype a complete physical/computational object. One goal is to suppress the challenges of mechanical design so that designers can focus on shape and usability. (3) Interactive machine learning techniques will be created to easily develop the mapping between human activities (as detected by sensors) and their recognition in software. It is known that putting humans in the training loop for machine learning changes the way training sets are built. Algorithms will be developed that both adapt to and exploit this behavior. This project will produce new knowledge of the concepts that are most difficult for designers when creating physical computational objects.
