Controlling Synthetic Characters in Simulations: A Case for Cognitive Architectures and Sigma

Simulations, along with other similar applications like virtual worlds and\nvideo games, require computational models of intelligence that generate\nrealistic and credible behavior for the participating synthetic characters.\nCognitive architectures, which are models of the fixed structure underlying\nintelligent behavior in both natural and artificial systems, provide a\nconceptually valid common basis, as evidenced by the current efforts towards a\nstandard model of the mind, to generate human-like intelligent behavior for\nthese synthetic characters. Sigma is a cognitive architecture and system that\nstrives to combine what has been learned from four decades of independent work\non symbolic cognitive architectures, probabilistic graphical models, and more\nrecently neural models, under its graphical architecture hypothesis. Sigma\nleverages an extended form of factor graphs towards a uniform grand unification\nof not only traditional cognitive capabilities but also key non-cognitive\naspects, creating unique opportunities for the construction of new kinds of\ncognitive models that possess a Theory-of-Mind and that are perceptual,\nautonomous, interactive, affective, and adaptive. In this paper, we will\nintroduce Sigma along with its diverse capabilities and then use three distinct\nproof-of-concept Sigma models to highlight combinations of these capabilities:\n(1) Distributional reinforcement learning models in; (2) A pair of adaptive and\ninteractive agent models that demonstrate rule-based, probabilistic, and social\nreasoning; and (3) A knowledge-free exploration model in which an agent\nleverages only architectural appraisal variables, namely attention and\ncuriosity, to locate an item while building up a map in a Unity environment.\n