Intelligent agents designed to perform in the real world should by definition be tested and evaluated in the real world. However, this is impossible in many situations: a lack of resources may rule out construction of a complete robotic environment, for example, or the desired domain may be physically inaccessible for testing. In such situations, the use of a simulation system to provide an environment in which to test and examine the intelligent system is necessitated. In the past, such systems have acquired a poor reputation within the AI community, mainly due to the sometimes grandiose claims of systems that are tested solely under simulated conditions. In this paper we explore the conditions under which simulation is justified, examine the inadequacies of currently available systems for the testing and examination of intelligent agents, and describe Gensim, a new system designed to address these inadequacies. Rather than providing a single, parameterized domain, Gensim provides a c...

Existing approaches to planning in Artificial Intelligence (such as Universal and Classical Planning) are designed for very specific types of activities, and are largely inapplicable to areas outside their narrow ranges. In particular, everyday activities that are simple for humans, such as making a meal or getting from place to place, require long-term goal-directed and timely responses that are far beyond the bounds of these traditional approaches. This dissertation examines the nature of the everyday activities and develops a computational architecture for an agent able to participate in such activities. An analysis of everyday activities shows them to be difficult tasks made artificially simple through extensive activity-specific knowledge possessed by the agent performing them. I argue that existing approaches are unsuitable to everyday activity because they rely too heavily on compiled knowledge and fail to adequately apply the background knowledge from which these compilations ...

Teleautonomy, the remote issuing of instructions for control of an agent that is intended to function semi-autonomously while employing those instructions, is an important issue in intelligent agency. In most cases where intelligent agents are employed, we neither wish to directly control every detail, nor trust them to deal completely autonomously with every important task. Similarly, in multi-agent settings, agents must similarly instruct and be instructed by other agents. This paper breaks the concept of Teleautonomy into a spectrum along several dimensions for research purposes, emphasizing the demands each dimension places upon an intelligent agent. We also describe two current approaches to teleautonomy in mobile agents that together cover a breadth of this range.

Intelligent agents designed to perform in the real world should, by definition be tested and evaluated in such a world. However, this is impossible in many situations: a lack of resources may rule out construction of a complete robotic environment, for example, or the desired domain may be physically inaccessible for testing. In such situations, the use of a simulation system is necessitated. In this paper we describe Gensim, a generic simulation system for intelligent agents. Rather than providing a single, parameterized domain, Gensim provides a collection of facilities allowing users to design complete environments for examining and testing intelligent agents. The system also provides direct low-level support for implementing sets of agents that display flexible autonomy: the ability to solve problems by varying the degree of autonomy of single agents in a multiagent system. 1. Simulation and Flexible Autonomy "At the intersection of computers, control, information, and managemen...