Adaptive Hierarchy of Distributed Fuzzy Control: Application to Behavior Control of Rovers

Created by W.Langdon from gp-bibliography.bib Revision:1.3872

@PhdThesis{tunstel:thesis,
  author =       "Edward W. Tunstel",
  title =        "Adaptive Hierarchy of Distributed Fuzzy Control:
                 Application to Behavior Control of Rovers",
  school =       "Electrical and Computer Engineering, University of New
                 Mexico",
  year =         "1996",
  address =      "Albuquerque, New Mexico, NM 87131, USA",
  month =        dec,
  keywords =     "genetic algorithms, genetic programming, robot",
  URL =          "http://www.cs.ucl.ac.uk/staff/W.Langdon/ftp/papers/TunstelPhD.pdf",
  URL =          "http://www.cs.ucl.ac.uk/staff/W.Langdon/ftp/papers/TunstelPhD.ps.gz",
  size =         "134+17 pages",
  abstract =     "This dissertation addresses the synthesis of
                 knowledge-based controllers for complex autonomous
                 systems that interact with the real world. A fuzzy
                 logic rule-based architecture is developed for
                 intelligent control of dynamic systems possessing a
                 significant degree of autonomy. It represents a novel
                 approach to controller synthesis which incorporates
                 fuzzy control theory into the framework of
                 behavior-based control. The controller intelligence is
                 distributed amongst a number of individual fuzzy logic
                 controllers and systems arranged in a hierarchical
                 structure such that system behaviour at any given level
                 is a function of behaviour at the level(s) below. This
                 structure addresses the combinatorial problem
                 associated with large rule-base cardinality, as the
                 totality of rules in the system are not processed
                 during any control cycle. A method of computationally
                 evolving fuzzy rule-bases is also introduced. It is
                 based on the genetic programming paradigm of
                 evolutionary computation and directly manipulates
                 linguistic terminology of the system. This provides a
                 systematic rule-base design method which is more direct
                 than current approaches that mandate numerical
                 encoding/decoding of rule representations. Finally, a
                 mechanism for multi-rule base coordination is devised
                 by generalisation of fuzzy logic theoretic concepts. It
                 is incorporated to endow the system with the capability
                 to dynamically adapt its control policy in response to
                 goals, internal system state, and perception of the
                 environment.

                 The validity and practical utility of the approach is
                 verified by application to autonomous navigation
                 control of wheeled mobile robots, or rovers. Simulated
                 and experimental navigation results produced by the
                 adaptive hierarchy of distributed fuzzy control are
                 reported. Results show that the proposed ideas can be
                 useful for realisation of autonomous rovers that are
                 meant to be deployed in dynamic and possibly
                 unstructured environments. This class of
                 computer-controlled, wheeled mobile vehicles includes
                 industrial mobile robots, automated guided vehicles,
                 office or hospital robots, and in some cases natural
                 terrain vehicles such as planetary rovers.

                 The proposed intelligent control architecture is
                 generally applicable to autonomous systems whose
                 overall behaviour can be decomposed into a bottom-up
                 hierarchy of increased behavioural complexity, or a
                 decentralised structure of multiple rule-bases.",
  notes =        "OCLC Number: 37306598. Author: Edward W. {Tunstel,
                 Jr.} Adviser Mohammad Jamshidi",
}

Genetic Programming entries for Edward W Tunstel

Citations