NASA Logo

NTRS

NTRS - NASA Technical Reports Server

Collections About News Help Login
Back to Results
Integrating planning and reactive controlArtificial intelligence research on planning is concerned with designing control systems that choose actions by manipulating explicit descriptions of the world state, the goal to be achieved, and the effects of elementary operations available to the system. Because planning shifts much of the burden of reasoning to the machine, it holds great appeal as a high-level programming method. Experience shows, however, that it cannot be used indiscriminately because even moderately rich languages for describing goals, states, and the elementary operators lead to computational inefficiencies that render the approach unsuitable for realistic applications. This inadequacy has spawned a recent wave of research on reactive control or situated activity in which control systems are modeled as reacting directly to the current situation rather than as reasoning about the future effects of alternative action sequences. While this research has confronted the issue of run-time tractability head on, in many cases it has done so by sacrificing the advantages of declarative planning techniques. Ways in which the two approaches can be unified are discussed. The authors begin by modeling reactive control systems as state machines that map a stream of sensory inputs to a stream of control outputs. These machines can be decomposed into two continuously active subsystems: the planner and the execution module. The planner computes a plan, which can be seen as a set of bits that control the behavior of the execution module. An important element of this work is the formulation of a precise semantic interpretation for the inputs and outputs of the planning system. They show that the distinction between planned and reactive behavior is largely in the eye of the beholder: systems that seem to compute explicit plans can be redescribed in situation-action terms and vice versa. They also discuss practical programming techniques that allow the advantages of declarative programming and guaranteed reactive response to be achieved simultaneously.
Document ID
19900019761
Acquisition Source
Legacy CDMS
Document Type
Conference Paper
Authors
Rosenschein, Stanley J. (Teleos Research Palo Alto, CA, United States)
Kaelbling, Leslie Pack (Teleos Research Palo Alto, CA, United States)
Date Acquired
September 6, 2013
Publication Date
January 31, 1989
Publication Information
Publication: JPL, California Inst. of Tech., Proceedings of the NASA Conference on Space Telerobotics, Volume 2
Subject Category
Administration And Management
Accession Number
90N29077
Funding Number(s)
CONTRACT_GRANT: NCC2-494
PROJECT: SU PROJ. 6359
Distribution Limits
Public
Copyright
Work of the US Gov. Public Use Permitted.

Related Records

There are no records associated with this record.
No Preview Available