Computing systems are embedded throughout our environment - in applications including "smart" appliances, environmental control systems, traffic and weather monitoring systems, and medical monitoring. Current techniques for controlling these devices employ fixed algorithms for sensory processing and for taking subsequent action. However, as these devices increase in their complexity, computing power, and sensory capabilities, it becomes possible to design systems that can automatically adapt their behavior to the specific environment and task in which they are embedded.
Undergraduate students involved in this multidisciplinary REU Site will investigate the use of machine learning techniques in the development of a variety of embedded systems. Application areas of focus include robot control; assistive, wearable computing systems; computer intrusion detection; and on-line weather prediction. Students will receive training in a variety of areas, including embedded system design, empirical methods for system evaluation, statistical machine learning techniques (including supervised and reinforcement learning, Bayesian methods, genetic algorithms and data mining), sensor processing, control, embedded interfaces, technical writing and oral presentation.
