Comp Sci: Machine Learning helping the elderly.

"Learning Machines"
O'Reilly Network (01/12/05); Oram, Andy

"Computer Professionals for Social Responsibility member Andy Oram believes robots could play a major role in boosting productivity for a rapidly aging population. But making robots capable of such a feat is a formidable challenge, as Carnegie Mellon University (CMU) research scientist Geoffrey Gordon indicated recently. The university devotes a lot of research to helping robots perceive their surroundings via distributed machine learning, a process that can be very complex because of the power requirements for robot-to-robot data transmission, as well as networked sensor nodes' fragility and susceptibility to interference in real-life environments; determining the most durable network scheme with the lowest transmission costs requires knowing the quality of links between nodes. Mesh networks--relatively looser systems with multiple redundant links--are widely hyped, but researcher Carlos Guestrin prefers a hierarchical tree architecture that delivers better scalability and gives each node a fairly accurate projection of environmental activity, though rapid self-reconfiguration is a must. Oram reports that Gordon appreciates commoditized technology for robots, and notes that CMU robots regularly use off-the-shelf Pentium chips running the Linux operating system. Gordon told Oram that predicting the emergence of major breakthroughs in robotics and artificial intelligence is difficult, because research often turns out to be a tougher prospect than originally assumed; "Progress tends to affect some deep aspect of the problem," Oram observes. The author details robotics research geared toward aiding disabled people, such as a sensor network that could perhaps detect a person in trouble and stage an intervention. Meanwhile, some research focuses on making robots capable of comprehending and responding to people's psychological states."

http://www.oreillynet.com/pub/wlg/6213