On approximate reasoning and minimal models for the development of robust outdoor vehicle navigation schemes

Conference · Mon Nov 01 04:00:00 EST 1993

OSTI ID:10192440

Pin, F G

Outdoor sensor-based operation of autonomous robots has revealed to be an extremely challenging problem, mainly because of the difficulties encountered when attempting to represent the many uncertainties which are always present in the real world. These uncertainties are primarily due to sensor imprecisions and unpredictability of the environment, i.e., lack of full knowledge of the environment characteristics and dynamics. Two basic principles, or philosophies, and their associated methodologies are proposed in an attempt to remedy some of these difficulties. The first principle is based on the concept of ``minimal model`` for accomplishing given tasks and proposes to utilize only the minimum level of information and precision necessary to accomplish elemental functions of complex tasks. This approach diverges completely from the direction taken by most artificial vision studies which conventionally call for crisp and detailed analysis of every available component in the perception data. The paper will first review the basic concepts of this approach and will discuss its pragmatic feasibility when embodied in a behaviorist framework. The second principle which is proposed deals with implicit representation of uncertainties using Fuzzy Set Theory-based approximations and approximate reasoning, rather than explicit (crisp) representation through calculation and conventional propagation techniques. A framework which merges these principles and approaches is presented, and its application to the problem of sensor-based outdoor navigation of a mobile robot is discussed. Results of navigation experiments with a real car in actual outdoor environments are also discussed to illustrate the feasibility of the overall concept.

• Research Organization: Oak Ridge National Lab., TN (United States)
• Sponsoring Organization: USDOE, Washington, DC (United States)
• DOE Contract Number: AC05-84OR21400
• OSTI ID: 10192440
• Report Number(s): CONF-9311107--1; ON: DE94002193
• Country of Publication: United States
• Language: English

Related Subjects

42 ENGINEERING
420203
COMPLEX TERRAIN
CONTROL SYSTEMS
FUZZY LOGIC
HANDLING EQUIPMENT AND PROCEDURES
MATHEMATICAL MODELS
MONITORS
NAVIGATION
ROBOTS