Quantitative modeling plays a key role in the natural sciences, and systems that address the task of inductive process modeling can assist researchers in explaining their data. In the past, such systems have been limited to data sets that recorded change over time, but many interesting problems involve both spatial and temporal dynamics. To meet this challenge, we introduce SCISM, an integrated intelligent system which solves the task of inducing process models that account for spatial and temporal variation. We also integrate SCISM with a constraint learning method to reduce computation during induction. Applications to ecological modeling demonstrate that each system fares well on the task, but that the enhanced system does so much faster than the baseline version.

Quantitative modeling plays a key role in the natural sciences, and systems that address the task of inductive process modeling can assist researchers in explaining their data. In the past, such systems have been limited to data sets that recorded change over time, but many interesting problems involve both spatial and temporal dynamics. To meet this challenge, we introduce SCISM, an integrated intelligent system which solves the task of inducing process models that account for spatial and temporal variation. We also integrate SCISM with a constraint learning method to reduce computation during induction. Applications to ecological modeling demonstrate that each system fares well on the task, but that the enhanced system does so much faster than the baseline version.

Although cognitive architectures provide an excellent infrastructure for research stretching over various fields, their integration of multiple modules makes their evaluation difficult. Due to the lack of analytical criteria, the cost of demonstrations, and varying specifications among different architectures, developing evaluation methods is a challenge. In this paper, we describe a testbed for evaluation that revolves around an in-city driving environment. With its familiar but challenging missions cast in a rich setting, the testbed provides a uniform and competitive environment for evaluating cognitive architectures.

This paper investigates a computational approach to transfer: the ability to use previously learned knowledge on related but distinct tasks. We study transfer in the context of an agent architecture, Icarus, and we claim that many forms of transfer follow automatically from its use of structured concepts and skills. We show that Icarus can acquire structured representations from domain experience, and subsequently transfer that knowledge into new tasks. We present results from multiple experiments in the Urban Combat Testbed, a simulated, real-time, three-dimensional environment with realistic dynamics.

While cognitive architectures provide excellent infras- tructure for research stretching over various fields, the integrated nature consisting of multiple modules makes their evaluation extremely difficult. Due to the lack of analytical criteria, the cost of general demonstra- tions, and varying specifications among different ar- chitectures, deriving any general evaluation methods is a complicated task. In this paper, we propose a method for empirical evaluation, using a testbed within an in-city driving environment. With its familiar but challenging missions cast in a rich setting, the testbed provides a uniform and competitive environment for agents, for evaluation of cognitive architectures that em- bodied them.

In this paper, we present a principled approach to constructing believable game players that relies on a cognitive architecture. The resulting agent is capable of playing the game \uc/ in a plausible manner when faced with similar situations as its human counterparts. We discuss how architectural features like goal-directed but reactive execution and incremental learning can produce more believable synthetic characters.

Skill execution has played a central role in cog- nitive architectures for managing various activ- ities. In simple domains, sequential skill exe- cution is usually sufficient to achieve a goal. However, in complex domains that involve multiple resources, serial skill execution can- not exploit all available resources. For this rea- son, an agent must have a mechanism to man- age resources for reasonable behaviors in these domains. We present an extension to teleo- reactive logic programs that supports parallel execution of multiple skills at the architectural level, and discuss preliminary evaluation of the new skill execution mechanism.