clarke@utcsri.UUCP (04/29/87)
[On account of a minor accident of otherwise very little interest, the regular notice for this week is unavailable.] (SF = Sandford Fleming Building, 10 King's College Road) (GB = Galbraith Building, 35 St. George Street) SUMMARY: SYSTEMS SEMINAR, Wednesday, May 6, 11 am, SF1101 -- Dr. Marc Abrams: ``Analysis of Distributed Programs With Periodic Equilibrium Behavior" GRAPHICS SEMINAR, Wednesday, May 6, 2 pm, GB120 -- Professor David Zeltzer: ``Motor problem solving for three dimensional computer animation" -------------------- SYSTEMS SEMINAR, Wednesday, May 6, 11 am, SF1101 Dr. Marc Abrams IBM Zurich Research Laboratory ``Analysis of Distributed Programs With Periodic Equilibrium Behavior" The state of many continuous physical systems (e.g., an electrical circuit) may reach a fixed value as time grows to infinity. Alternately, the system may reach a limit cycle behavior, in which the stable system state varies periodically in time. Certain distributed programs display similar behavior. During execution, they quickly enter one of several periodic or oscillating modes. The mode reached depends on the process starting order, communication delay, and the structure and timing of pro- cess synchronization. This talk describes a novel performance analysis technique, the geometric concurrency model. It predicts the sequence of synchronization points where a program blocks, the blocking durations, and the duration of concurrent execution between synchronization points. We then discuss use of the technique to predict the behavior of a dining phi- losophers program on the ZMob distributed computer. Finally, we explain why knowledge of periodic equilibrium behavior is important in programming dis- tributed programs. GRAPHICS SEMINAR, Wednesday, May 6, 2 pm, GB120 Professor David Zeltzer Massachusetts Institute of Technology ``Motor problem solving for three dimensional computer animation" In the recent past we have seen great improvements in the ability to generate and display complex synthetic imagery. Near-photographic realism can now be achieved for certain classes of objects and landscapes. However, current computer animation systems require far too much technical or programming experience on the part of the user, so that the power of computer-based simulation and animation is gen- erally unavailable to professionals who could make use of such tools for educa- tion, design and research. I argue that this is not simply a user inter- face issue. Rather, we need to incorporate a rich representation of the task domain, so that users can describe an animated sequence as a set of events and relationships involving the characters and objects in a scene. I call this TASK LEVEL animation. However, even the simplest scenes and actions involving moving fig- ures can require solving a host of seemingly trivial difficulties. All kinds of animals as well as human beings have the capacity to move and function in the physi- cal world efficiently, in the face of changing conditions, apparently without conscious intervention. A fundamental problem of task level animation, there- fore, is to simulate the routine and stereotypical -- yet higly adaptive -- behaviors of agents in virtual microworlds. I describe a theory of MOTOR PROBLEM SOLVING for task level animation of human and animal figures in simulated environments. I propose that a simple but very general problem solving capacity is innate in the organiza- tion of an agent's behavior repertoire, using a strategy of local means-ends analysis for forward-chaining through a lattice of motor skills without back- tracking. The domain of this problem solving paradigm is characterized, we describe its failure modes, and suggest that advice-taking and mechanisms of attention and graded recruitment of resources are plausible means of recovery from failure. -- Jim Clarke -- Dept. of Computer Science, Univ. of Toronto, Canada M5S 1A4 (416) 978-4058 {allegra,cornell,decvax,linus,utzoo}!utcsri!clarke