call for papers, previous message From: hexmoor@cs.buffalo.edu (henry h hexmoor) Subject: CFP: 1995 AAAI Spring Sympoisum Date: Wed, 27 Jul 1994 02:37:11 GMT 1995 AAAI Spring Symposium Date: March 27-29, 1995 Place: Stanford, CA Deadline for submission: October 28, 1994 URL: http://tommy.jsc.nasa.gov/er/er6/mrl/symposium.html --------------- Lessons Learned from Implemented Software Architectures for Physical Agents =========================================================================== We are interested in organizational concepts for artificial agents that function in the everyday world of manufacturing floors, office buildings, and houses or in the specialized worlds of space or nuclear reactors. In recent years, a sufficient number of researchers have put forth software frameworks for organizing intelligence in agents beyond the original first few who began such investigations [e.g., SOAR and NASREM] that a symposium to discuss the issues outlined below is warranted. We are seeing architectures with as few as three and as many as five layers of cognition or control, for single agents or multiple agents, designed to accommodate hard real-time constraints or involved user interfaces, handling purely reactive or a combination of deliberate and reactive control, which are subsumptive or supervenient, and most of which are designed to function as part of a physical agent. The goal of this workshop is to shed light into reasons for architectural decisions in building artificial agents. Many important questions affect architectural decisions. For this workshop, we ask the following questions only with respect to architectural decisions. Coordination-- How should the agent arbitrate/coordinate/cooperate its behaviors and actions? Is there a need for central behavior coordination? Interfaces-- How can human expertise be easily brought into an agent's decisions? Will the agent need to translate natural language internally before it can interact with the world? How should an agent capture mission intentions or integrate various levels of autonomy or shared control? Can restricted vocabularies be learned and shared by agents operating in the same environment? Representation-- How much internal representation of knowledge and skills is needed? How should the agent organize and represent its internal knowledge and skills? Is more than one representational formalism needed? Structural-- How should the computational capabilities of an agent be divided, structured, and interconnected? What is the best decomposition/granularity of architectural components? What is gained by using a monolithic architecture versus a multi-level, distributed, or massively parallel architecture? Are embodied semantics important and how should they be implemented? How much does each level/component of an agent architecture have to know about the other levels/components? Performance-- What types of performance goals and metrics can realistically be used for agents operating in dynamic, uncertain, and even actively hostile environments? How can an architecture make guarantees about its performance with respect to the time-critical aspect of the agent's physical environment? What are the performance criteria for deciding what activities take place in each level/component of the architecture? Psychology-- Why should we build agents that mimic anthropomorphic functionalities? How far can/should we draw metaphoric similarities to human/animal psychology? How much should memory organization depend on human/animal psychology? Simulation-- What, if any, role can advanced simulation technology play in developing and verifying modules and/or systems? Can we have standard virtual components/test environments that everybody trusts and can play a role in comparing systems to each other? How far can development of modules profitably proceed before they should be grounded in a working system? How is the architecture affected by its expected environment and its actual embodiment? Learning-- How can a given architecture support learning? How can knowledge and skills be moved between different layers of an agent architecture? We invite researchers in intelligent mobile robots, robot manipulators, autonomous creatures (animats), and neuroscience as applied to autonomous agents to join us in discussing these questions. To allow for a more practical discussion of the issues, all submissions should focus on an agent or agents performing a specific task, such as keeping a house clean, maintaining the space station, or delivering parts on a factory floor. Be very specific about how your agent(s) organize(s) its knowledge and skills in order to perform this task and what mechanisms your agent(s) use(s) to invoke the correct knowledge or skill at the appropriate time. Please include the design decisions you made in organizing your agent's architecture for the task. Then, through this specific example, show how your agent's architecture addresses some of the questions listed above. The symposium will consist of presentations, invited talks, and task groups. Based on submissions, we will divide the workshop into specific task groups and, after discussions, come together for synthesis. We are tentatively proposing that the group produce as a minimum a set of answers for a portion of the discussion areas listed above. SUBMISSION INFORMATION Potential attendees should submit either an extended abstract or a full paper, neither of which should exceed 20 pages. If you are submitting a paper, we prefer that it not have been published elsewhere. If you are sending a paper that has already been published, tell us where it appeared. We encourage everyone to submit their papers or abstracts electronically, PostScript or ASCII only. Submission can be made by either e-mailing the entire document or e-mailing an anonymous ftp address for the document. E-mail submissions should be sent to: hexmoor@cs.buffalo.edu If e-mail submission is not possible, please send three copies of the paper or abstract to: Henry Hexmoor Co-chair, AAAI Spring Symposium 226 Bell Hall Dept of Computer Science SUNY at Buffalo Buffalo, NY 14260 ORGANIZING COMMITTEE Ron Arkin Peter Bonasso Henry Hexmoor (co-chair) David Kortenkamp (co-chair) David Musliner GUEST SPEAKERS James Albus George Bekey Mike Brady -- No man is an Island, entire of it self; every man is a piece of the Continent, a part of the main; if a clod be washed away by the sea, Europe is the less, as well as if a promontory were, as well as if a manor of thy friends or of thine own were; any man's death diminishes