MV4015 Agent-Based Autonomous Behavior for Simulations

Covers the concepts and skills required to apply agent-based programming to models and simulations of complex adaptive systems (CAS). Concepts covered will include: complex systems - especially their properties of path dependence, sensitivity to initial conditions, emergence of self-organized structure, adaptation to a changing environment, and criteria for evaluation model or simulation fidelity; distinctions between agent-based methods and other kinds of programming; goal-directed behavior and decision making; situational reasoning and the elements of rational behavior. The course will survey specific works and key contributors to this subject: John Holland, Complexity Science at the Santa Fe Institute, Artificial Life, Brian Arthur (the El Farol Problem and Bounded Rationality), SWARM, Sugarscape, ISAAC, Daniel Dennett (Intentionality), Richard Dawkins. Within this conceptual and historical framework the course will emphasize design, specification, and programming skills that will equip the student to know when and how to apply agent-based methods to models and simulations. The programming skills will involve genetic algorithms, classifier systems, applications of game theory and reinforcement learning, and the treatment of collaboration and defection in groups. Finally, the course will discuss agent-based simulations in the context of distributed, virtual environments. Prerequisites: None.

Lecture Hours

4

Lab Hours

2

Quarter Offered

  • As Required