Brief Overview of Cognitive Agents for Social Environments (CASE)
A general abstract of the goals of the CASE system can be found here. In regard to creating a simulation on this system, CASE works in the following ways.
The construction of a simulation with CASE begins with the interface control. From this graphic interface, a user may select various options for the system, such as how many and which computers to use, which simulation file to use, how many steps or iterations to simulate, and any options specific to the selected simulation (e.g., number of agents, size of environment, etc.).
The master server is then started on the local machine, which initializes rudimentary parameters such as the bounds of the simulation and a manager which will handle any services you try to connect to the simulation, such as SwiftVis. Additionally, if agents cannot find other agents when communicating, the manager will register all agents, as a backup, so they can be found later.
The master server then determines how to divide the simulation across the selected machines. The master server opens a connection to each machine to start a slave server. Each slave server then registers all the agents that are currently on it. As the agents begin to communicate to agents on other machines, links to these remote agents are registered as well.
When initialization is complete, the master server keeps the slave servers synchronized as the slave servers work through each pase of each step of the simulation. As the bounds for each slave server change (in accordance with load balancing) and as agents move across bounds, the agents are appropriately unregistered from their original slave server and registered onto the new slave server.
All agents must have a unique ID, so whenever agents want to talk to other agents, the servers can find the correct agent by this ID. This search begins at the slave servers and works up to the master server if the ID cannot be found.
Note: In the current state of the system, changing the state of an agent through communication with another agent is not allowed, as communication is intended strictly for information retrieval purposes.
Next, take an overview of the Agent and Sugar Classes.
Back to Main Menu
