Modeling and Simulation Work Practice with the Brahms Agent Environment

Half-day tutorial (1:00 - 4:30pm) [more information on Brahms]

Maarten Sierhuis, Ph.D.

More and more people are interested in developing "day in the life" models and simulations of people's behavior at the second and longer timeframe, the interaction between groups of people and systems, as well as the movement and interaction with the environment. Cognitive modeling tools (e.g., Soar, ACT-R) focus on detailed modeling of individual cognitive tasks at the sub-second level. In contrast, activity modeling focuses on higher-abstraction behaviors that enable modeling of people's daily activities and enable a focus on how informal, circumstantial, and located behaviors of a group of individuals occur and where communication and synchronization happen, such that the task contributions of people and machines flow together to accomplish goals. This is referred to as "work practice modeling." The tutorial will provide an overview of the Brahms multi-agent activity modeling language by considering a simple 'day in the life' scenario. There will be hands-on experience with Brahms.

Brahms includes an activity-oriented Belief-Desire-Intention (BDI) language, a compiler and virtual machine for executing Brahms models, as well as an Eclipse plug-in and a post-execution viewer of agent execution, communication and interaction. Brahms enables the creation of multi-agent models that include aspects of reasoning found in cognitive models, task execution, plus the impact of interaction and geography, such as agent movement and physical changes in the environment. Brahms is currently used to automate the work of a flight controller in NASA's International Space Station's Mission Control Center (ISS MCC). This system, called OCAMS, has been in production in the ISS MCC, 24x7, since July of 2008, and is based on a Brahms model of the work practices of the flight controllers. OCAMS is a distributed Multi-Agent System.

Prerequisite knowledge: There is no prerequisite for taking this tutorial. A useful background to have is some experience in agent architectures, especially belief-desire-intention architectures and discrete-event simulation.

Maarten Sierhuis recently joined PARC as area lead for the new Knowledge, Language and Interaction area. Before this, he was at NASA Ames for over twelve years, working on and applying Brahms at NASA. He is a Co-Principal Investigator for the Brahms multi-agent environment. He is also a visiting professor at the Man-Machine Interaction group at Delft University of Technology, where he teaches a graduate course using Brahms, called Agent-Based Modeling and Simulation of Organizations and Work Practice. He has a Ph.D. in Social Science Informatics from the University of Amsterdam and an engineering degree in Informatics from the Polytechnic University in The Hague, The Netherlands. He has presented invited lectures and tutorials on both Brahms and Compendium, and has published widely in these areas.