The Role of Chess in Artificial Intelligence Research

Levinson, Robert; Hsu, Feng-Hsiung; Schaeffer, Jonathan; Marsland, T.A.; Wilkins, David E.

doi:10.3233/icg-1991-14314

Cited by 7 publications

(7 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Many planning problems have a fully observable state space, such as the game of chess, where tree search methods can compute excellent move plans [1]. Even planning problems that have a large state space, often have limited action sets to choose from, such as backgammon, where existing approaches solve these tasks well [2].…”

Section: Imentioning

confidence: 99%

“…RISK is a non-cooperative stochastic game model with multiple players, where searching deeply for viable alternatives weighs strongly in game play [10] [11], much as in chess [12]. The work presented here is an extension of this approach for partially observable games.…”

Section: Introductionmentioning

confidence: 98%

“…Having both a large unknown state space and a large set of available actions presents the opportunity for new approaches to dynamic planning. 1 Having known state transitions implies that one can quantify an action's effect on the environment. Thus, given an action, the number of possible outcomes is finite and quantifiable as a state transition.…”

mentioning

confidence: 99%

See 2 more Smart Citations

Automated dynamic planning and execution for a partially observable game model: Tsunami City search and rescue

Vaccaro

Guest

2008

2008 IEEE Congress on Evolutionary Computation (IEEE World Congress on Computational Intelligence)

View full text Add to dashboard Cite

This paper addresses the problem of autonomous dynamic planning and execution (ADP&E) for partially observable model environments. There are three accomplish-ments illustrated in this paper: (1) develop an ADP&E implementation framework for planning and executing in partially observable model environments, (2) design and implement a methodology for adapting planner parameters to improve the overall planning process, and (3) demonstrate the utility of the planning process on a large complex application (i.e., city search and rescue operations). I.PROBLEM DOMAIN he application implemented for this paper demonstrates the following accomplishments: (1) an ADP&E framework for planning in partially observable environments, (2) a method for adapting planning parameters, and (3) an application of planning a search and rescue (S&R) operation. The S&R problem includes a partially observable environment and the planning system of concern here is one with the following attributes:Very large state-space representation (> 10 100 states) Large finite set of available actions (> 10 3 ) Known state transitions (possibly stochastic) Finite set of measurable features and goals Having a large, partially hidden state space and a large finite set of available actions favor using the approach described in this paper. Many planning problems have a fully observable state space, such as the game of chess, where tree search methods can compute excellent move plans [1]. Even planning problems that have a large state space, often have limited action sets to choose from, such as backgammon, where existing approaches solve these tasks well [2]. Having both a large unknown state space and a large set of available actions presents the opportunity for new approaches to dynamic planning. 1 Having known state transitions implies that one can quantify an action's effect on the environment. Thus, given an action, the number of possible outcomes is finite and quantifiable as a state transition. Also, having a finite set of quantifiable features implies that there is a portion of the state-space that can be observed and quantified for any alternate outcome at any given time. These quantities can then be evaluated to determine the possible action choices that bring about the more desirable outcomes. A finite set of measurable goals implies that there are metrics that determine action and/or plan selection. The goal metrics are not prescribed as fixed rules or stateaction pairs, but parameterize a selected group of features into an adaptive form. Specifically, each feature is considered a unique dimension in the space, while each goal corresponds to features' desirability.

show abstract

Section: Imentioning

confidence: 99%

Section: Introductionmentioning

confidence: 98%

See 1 more Smart Citation

Automated dynamic planning and execution for a partially observable game model: Tsunami City search and rescue

Vaccaro

Guest

2008

2008 IEEE Congress on Evolutionary Computation (IEEE World Congress on Computational Intelligence)

View full text Add to dashboard Cite

show abstract

“…Carefully engineered, deep-searching computers now dominate all but a few humans in a number of challenging games, including chess, checkers, and Othello. The surprising success of the engineering approach on these games has prompted researchers in other fields to seek similar search-intensive solutions to their problems, including theorem proving and natural language processing (see Marsland's discussion in (Levinson et al 1991)).…”

Section: Introductionmentioning

confidence: 99%

“…The preceding discussion illustrates that, even in the wake of major successes, the field of computer game playing is reevaluating itself. This introspection was manifested in a successful IJCAI-91 panel on the role of computer chess in A1 research (Levinson et al 1991). More recently, a growing community of A1 researchers is interested in addressing the limitations of the traditional approach from both scientific and engineering perspectives.…”

Section: Introductionmentioning

confidence: 99%