ProtoAndroid Progress Report #5

David Santos
Robot Group

ERROR- Lost Formatting

Robofest 5 was a public debut for two key features of the PA project. Invited speakers Joanne Pransky (Robot's "Psychiatrist") and Steve Ryan (Robotics Lawyer) gave a lively talk about a scheme to endow PA with a legal and social identity. Steve has elaborated a plan whereby a corporation is formed with the 'droid on the board and the human boardmembers resign in favor of the entity, while retaining contractual guardianship as necessary. A small expert system which Steve is writing will give PA the ability to manage the sale of its own stock in accordance with need for maintenance and upgrades. PA will have a small endowment to manage and may derive additional income as a celebrity speaker. The other big news was the unveiling of a realtime photo-realistic FaceSpace (0.3) program that gives intelligent agents like PA a facial vocabulary tied to system and proximity events. This is "Max Headroom" technology. While the excellent graphics running on a Pentium are the most visible element, the heart of the system is a clean multi dimensional state space matrix with a semantic interface. Over 36,000 different facial expressions are possible due to a clever compression stategy that achieves about 800 to 1 reduction in needed storage. Instead of gigabytes of stored graphics FaceSpace needs only 2 megs to contain a 24 bit facial agent. Internal components include a library of every emotion we can think of and a set of rules that determine dynamic behavior based on system or proximity events. Tim O'Leary and Gilbert Andrade have done quite a bit of work to get FaceSpace to its present state of operability. Mike Wren provided key financial and moral support. He made the Pentium purchase possible thru Faustex Systems Corp. John Witham has provided critical media services thru Eyecon Interactive Media. Wan Yik-Lee has proposed programming an ApeBoard based FaceSpace compatible animatronic head. Brooks Coleman and Jerald Knight said they would be interested in its fabrication (Jerald hosted an animatronic technology table at Robofest and is developing shells and skin for PA). A second generation proto-android is being studied and will feature ultralight composite construction (a spin-off of Brooks' and my work on the Martial Arts Robot) and hyper- realistic appearance. The new head design might logically be that of the new machine. Work continues in the construction of PA's carcass. The skeleton is 2/3 complete and donations of necessary but expensive components such as pneumatic cylinders and valves are accumulating. There was disappointment that the robot was static at Robofest. The decision was made not to spend large amounts of time rigging up motion only to tear down the machine in order to continue fabrication. Motion of key components was demoed at Robofest 4 and video is available for viewing. Jerry Burton, president of the Southern California Robotics Society, showed great interest in the conceptual models and large computer generated system schematic diagrams being used to define PA's configuration. Many of you have seen early versions of this stuff over the years but may be surprised at the progress. Paco made a fleeting reference to one of the wall sized scratch sheets in Mondo 2000. A forth coming report will describe these tools in detail. Liz Stice of National Instruments sends regards from the maternity ward. Our donation request is still in process and looks good although there have been delays unrelated to its merits. Donations are proceeding nicely from Austin Integrated Systems. Bimba's old commitment to give cylinders needs to be renewed because installation is pending. At the next Robot Group meeting I'll propose that the ProtoAndroid project serve as the first test of a model Robot Group project review process. Because we receive public funding under a sort of sacred trust it is important that projects be carefully followed to maintain high performance. Besides serving as a first benchmark, such a review of Pa's evolution might serve as a useful "marketing tool" in that favorable findings will be supported independent of claims made by internal reports. I only wish there were actual competing android projects to compare with. Somebody please start one or let us know who has. Buzz term of the month- Persistent Agents. Persistence is defined as a mostly continuous state of on-ness and productive activity in an intelligent autonomous system. Other important agent qualities include breadth, or richness of behavior, and interactivity, the ability to engage directly with other agents and the environment. The easy part of fielding a persistent agent is leaving the power switched on, the harder part is to avoid having behavior wind into a repetative rut. Please let me know if you want detailed information on any of the topics in these reports.

Thanks for flying PolyCosmos. Join us again.