Action and planning in embedded agents

Kaelbling, Leslie Pack; Rosenschein, Stanley J.

doi:10.1016/s0921-8890(05)80027-2

Cited by 114 publications

(36 citation statements)

References 1 publication

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…After the advice has been parsed, the system transforms the general advice into terms that it can directly understand. Using techniques from knowledge compilation (Dietterich, 1991), a learner can convert ("operationalize") high-level advice into a (usually larger) collection of directly interpretable statements (Gordon & Subramanian, 1994;Kaelbling & Rosenschein, 1990;Nilsson, 1994). We only address a limited form of operationalization, namely the concretization of imprecise terms such as "near" and "many."…”

Section: A General Framework For Advice-takingmentioning

confidence: 99%

Creating advice-taking reinforcement learners

Maclin

Shavlik

1996

Mach Learn

View full text Add to dashboard Cite

Abstract. Learning from reinforcements is a promising approach for creating intelligent agents. However, reinforcement learning usually requires a large number of training episodes. We present and evaluate a design that addresses this shortcoming by allowing a connectionist Q-learner to accept advice given, at any time and in a natural manner, by an external observer. In our approach, the advice-giver watches the learner and occasionally makes suggestions, expressed as instructions in a simple imperative programming language_ Based on techniques from knowledge-based neural networks, we insert these programs directly into the agent's utility function. Subsequent reinforcement learning further integrates and refines the advice. We present empirical evidence that investigates several aspects of our approach and shows that, given good advice, a learner can achieve statistically significant gains in expected reward. A second experiment shows that advice improves the expected reward regardless of the stage of training at which it is given, while another study demonstrates that subsequent advice can result in further gains in reward. Finally, we present experimental results that indicate our method is more powerful than a naive technique for making use of advice.

show abstract

Section: A General Framework For Advice-takingmentioning

confidence: 99%

Creating advice-taking reinforcement learners

Maclin

Shavlik

1996

Mach Learn

View full text Add to dashboard Cite

show abstract

“…-the agent network architecture developed by Pattie Maes [32,34,35]; -Nilsson's teleo reactive programs [40]; -Rosenchein and Kaelbling's situated automata approach, which is particularly interesting in that it shows how agents can be specified in an abstract, logical framework, and compiled into equivalent, but computationally very simple machines [47,28,27,48]; -Agre and Chapman's pengi system [1]; -Schoppers' universal plans -which are essentially decision trees that can be used to efficiently determine an appropriate action in any situation [52]; -Firby's reactive action packages [17].…”

Section: The Objective Is To Explore a Distant Planet More Concretelmentioning

confidence: 99%

Intelligent Agents: The Key Concepts

Wooldridge

2002

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. This chapter aims to introduce the reader to the basic issues surrounding the design and implementation of intelligent agents. It begins by motivating the idea of an agent, presents a definition of agents and intelligent agents. The article then goes on to discuss four major approaches to building agents. First, logic based architectures are reviewed, in which decision-making is viewed as logical deduction. Second, reactive architectures are discussed, in which symbolic representations and models are eschewed in favour of a closer relationship between agent perception and action. Third, we discuss belief-desire-intention architectures, in which decision making is viewed as practical reasoning from beliefs about how the world is and will be to the options available to an agent, and finally to intentions and actions. Fourth, we review layered agent architectures, in which decision making is partitioned into a number of different decision making layers, each dealing with the agent's environment at a different level of abstraction.

show abstract

“…Another sophisticated approach is that of Rosenschein and Kaelbling [110,111,76]. In their situated automata paradigm, an agent is specified in terms of a logic of knowledge.…”

Section: Rosenschein and Kaelbling -Situated Automatamentioning

confidence: 99%