Behavior generation in intelligent systems
James S. Albus, Alexander Meystel
- Year
- 1997
- Citations
- 15
- Access
- Open access
Abstract
This book is about architectures for behavior generation.One should distinguish the concept of "architecture" from the concept of "algorithm".The same architecture can employ different algorithms while the same algorithm can be applied in many architectures.Architecture studies how various functional components should be put together to provide for the desirable functioning of a system.Architecture contemplates the need in partucular components and their interrelatedness if a concrete functioning is required.In other words, architecture can be considered a "metalogic" of the system, or its "meta-algorithm".This book was stimulated by the problem of developing Intelligent Systems in the variety of application domains such as robotics, integrated manufacturing, large complex systems in costruction.The techniques we describe are to be applied for design and control of all complex integrated technologies where NIST-RCS Architecture is appropriate.It should be considered further development of the general NIST-RCS concept.The thrust of this book is the concept of a recursive BG-module and the concept of multiresolutional, nested Loops of Functioning.The material contains main statements and positions of the 1992-1994 discussions between J. Albus and A. Meystel.The materials of these discussions have been developed and revised during the period from November 1994 through December 1995.NIST-RCS** has several versions created resulting from gradual evolution of the initial RCS concept within the NASREM* architecture to the current version.Although the NIST-RCS architecture has many successful applications, they are only particular solutions.This book elaborates particularly upon one of the NIST-RCS subsystems: the subsystem of "Behavior Generation" (BG).The BG subsystem is the driving force for the overall functioning of the NIST-RCS architecture.It propagates Goals top down and requests for the goal corrections from the bottom up.This provides for the transformations from the general Goals, Subgoals, Assignment, and Tasks into plans.These include Job Assignments and Schedules, feedforward and feedback control laws as well as the sequences of control commands.The system of Behavior Generation is * "NASREM" is the NASA/NBS Standard Reference Model for the Space Station Telerobotics [14].This was one of the first descriptions of the mutiresolutional hierarchy of organization proposed for analysis and design of complex systems.**The term "RCS" is partially misleading because of the bulk of literature dedicated to a variety of real- time operating computer-based controllers (see, for example, Eds.J. Stankovic, K. Ramamritham, Advances in Real-Time Systems, IEEE CS Press, 1993).In the meantime, in the last meeting on Hybrid Controllers, most of the authors spoke about the results obtained at NIST referring to them as to "NASREM" hierarchical controller.This is why the term NIST-RCS seems to reflect the entire situation in the area.the hub for the NIST-RCS architecture functioning, and for using NIST-RCS architecture practically.The BG-problem should be resolved in its entirety.But Behavior Generation is more than Planning/Control activities.It would be impossible without constant use of the World Modela representation of the World which is valid for the interval of time which can be associated with the word "Now" at each of the hierarchical levels.Actually, Behavior Generation is the result of interacting of the Planning/Control system with the World Model and exploring how and what can be anticipated by the World Model concerned with the decisions the system is about to make.In their discussions, J. Albus and A. Meystel addressed a refined structure of the "Behavior Generation" subsystem within a conceptual paradigm which can reduce difficulties of task decomposition including planning and control.This material also contains the results of discussions with members of the Architecture Group of Intelligent Systems Division (ISD) and Manufac
Keywords
Related papers
Statistical Learning Theory
Yuhai Wu, Vladimir Vapnik
1999
Artificial intelligence: a modern approach
1995
Fractional Differential Equations
Igor Podlubný
2025
Applied Nonlinear Control
Jean-Jacques Slotine, Weiping Li
1991