Agents in dynamic contexts, a system for learning plans

tells the story of two aritificial agents -David and Teddy. While David is indistinguishable from a human child, his companion Teddy is much simpler. Its behavior, however, still suggests a crucial mix of capabilities that stretch the state of the art in AI today. We argue that (unlike most contemporary AI), Teddy qualifies as a bona fide agent and that implementing such a system would represent a valuable advance in our understanding of agency. We then describe a project to integrate our existing work to create a simple agent with Teddy-like capabilities.

Inspired by the theory of practical reasoning, Belief-Desire-Intention (BDI) agents are perhaps the most well-known type and architecture of cogni-tive agents. Such agents can reason about their environment and perform complex plans to bring about their objectives and goals. Within the context of ever-changing environments though, one desirable feature for agents is that of learning , implemented in BDI agents as Intentional Learning, a framework focused on the monitoring of the mental states to include learning as part of the agent goals. In this paper, we consider and develop intentional learning within the Jason BDI framework for agents focused on a plan acquisition strategy addressing the cases of learning plans composed of one action, sequences or a repetition of actions that allow an agent to improve its behaviour at run-time. This is done at the pure BDI agent level, the repertoire of plans is directly updated without using external planning tools. We take as a testbed the simple vacuum cleaning environment and how new plans are acquired for accomplishing tasks of different level of complexity: escape from tunnel-like paths and wall-following. Furthermore, we integrate in a novel way the use of NetLogo as an environment to locate Jason agents, maintaining a clear delineation between decision making and action in the environment with the decision-making firmly anchored within the BDI agent's reasoning cycle.

2006

1995

This paper presents an approach to learning from situated, interactive tutorial instruction within an ongoing agent. Tutorial instruction is a exible (and thus powerful) paradigm for teaching tasks because it allows an instructor to communicate whatever types of knowledge an agent might need in whatever situations might arise. To support this exibility, however, the agent must be able to learn multiple kinds of knowledge from a broad range of instructional interactions. Our approach, called situated explanation, achieves such learning through a combination of analytic and inductive techniques. It combines a form of explanation-based learning that is situated for each instruction with a full suite of contextually guided responses to incomplete explanations. The approach is implemented in an agent called Instructo-Soar that learns hierarchies of new tasks and other domain knowledge from interactive natural language instructions. Instructo-Soar meets three key requirements of exible instructability that distinguish it from previous systems: (1) it can take known or unknown commands at any instruction point; (2) it can handle instructions that apply to either its current situation or to a hypothetical situation speci ed in language (as in, for instance, conditional instructions); and (3) it can learn, from instructions, each class of knowledge it uses to perform tasks.

Interface agents are computer programs that employ Artificial Intelligence techniques in order to provide assistance to a user dealing with a particular computer application. The paper discusses an interface agent which has been modelled closely after the metaphor of a personal assistant. The agent learns how to assist the user by (i) observing the user's actions and imitating them, (ii) receiving user feedback when it takes wrong actions and (iii) being trained by the user on the basis of hypothetical examples. The paper discusses how this learning agent was implemented using memory-based learning and reinforcement learning techniques. It presents actual results from two prototype agents built using these techniques: one for a meeting scheduling application and one for electronic mail. It argues that the machine learning approach to building interface agents is a feasible one which has several advantages over other approaches: it provides a customized and adaptive solution which is less costly and ensures better user acceptability. The paper also argues what the advantages are of the particular learning techniques used.

Computers & Education

The use of plan recognition in the user interface is presented as a basis for a learning environment within which students can assimilate the range of possible actions and objectives afforded by a target application. Support in the form of "present whereabouts" and "possible progressions" enable students to learn by doing. When combined with a facility for exploratory learning, this helps avoid the onerous air of a '*teaching situation".

1997

This paper describes a software agent that learns procedural knowledge from a human instructor well enough to teach human students. In order to teach, the agent needs more than the ability to perform a procedure. It must also be able to monitor human students performing the procedure and be able to articulate the reasons why actions are necessary. Our research concentrates on helping an instructor instruct the agent in a natural manner, on reducing the burden on the instructor, and on focusing learning on the procedure being taught. Initially the agent has little domain knowledge. The instructor demonstrates a procedure by directly manipulating a simulated environment. However, one demonstration is not sufficient for understanding the causal relationships between a demonstration's actions. Unfortunately, the more demonstrations a procedure requires, the greater the instructor's burden. However, fewer demonstrations can be required if the agent autonomously experiments. Our experiments attempt to understand the causal dependencies between a demonstration's actions by perturbing the order of the demonstration's actions.

1999

ACKNOWLEDGEMENTS This research was successfully completed thanks to many persons who helped me at various stages. I thank Dr. John Bourne who, as my advisor, coached me through the steps to completion. I thank my other outstanding committee members, Dr. Arthur Broderson, Dr. Alrajeh, and John Crocetti who supported and helped making this research project possible. The success of this research effort is directly attributable to the support and cooperation of the ALN Center staff. I am thankful for my father and mother for their unfailing supports, encouragement and inspiration throughout my doctoral education at Vanderbilt University.

ubicc.org

The world itself is a very complex place for architecture agents to act in. The real world is partially observable, stochastic, sequential, dynamic and continuous. Most architecture is designed to only deal with fractions of the total possible environmental ...

The confluence of new technologies and the increasing adoption of standards are pushing eLearning to the forefront of web-based initiatives. Agent technology is seen as a critical factor in knowledge acquisition in semantically rich environments, as exemplified by the Semantic Web. In addition the development and delivery of learning content is facilitated by the adoption of eLearning standards. In this paper we consider an eLearning framework where the learning process is mediated by software agents and supported by a knowledge base on learner profiles and skill maps. The roles of the agents are closely associated with the different stages of the learning process. The proposed approach is put into perspective through a brief discussion of some of the challenges that eLearning development is facing.