Artificial Agents
29
Swarm is an example modeling framework that works as an event-based
model. In time-driven approach, the system is updated at end of a function map. FLAME works on a time-driven approach with synchronous
updating where all agents are updated at same time and in parallel.
Asynchronous updates in a model take place when agents are updated
in random order or based on their own internal clocks.
Distributed. Multi-agent systems are concerned with distributed and coordinated problem solving. Bond and Gasser [26] describe distributed AI
in three areas:
• Distributed problem solving (DPS): how a problem is divided
among a number of nodes to be solved in parallel using knowledge
sharing.
• Multi-agent systems (MAS): concerned with ‘coordinating intelligent behavior among a collection of autonomous intelligent agents’.
• Parallel AI (PAI): concerned with performance like different computational speeds and finding new paths for problem solving.
Some agent-based modeling frameworks use CNET protocol, which work
on principle of a manager managing a set of workers. Every task is
decomposed into smaller subtasks and suitable nodes are selected to
work on the sub-task. At the end, the results are then integrated together
for a complete solution.
Decentralized behavior. Complex systems are decentralized and individuals make decisions based on their locations. Each agent evolves depending on information received locally. Whereas, evolution is based
on private memory and messages. Over time niches form, where some
agents do better than others.
Messaging. This is an important aspect of agent-based models allowing communication between agents. These interactions are responsible for emergent behavior. This follows the distributed nature of agent-based models, where messaging ensures all messages are read before decisions are
made.
Parallelism in agents. Some agent production systems use if-then statements to update rules. These rules determine the next state moved to.
A knowledge database is plugged into resolve and execute the rules.
Various parallel AI languages, like Prolog, can be used to code these
examples. However, it is important to parallelize work and synchronize
among all agents to share information. Example factors considered with
parallelism are
• Task parallelism
• Match parallelism
29
Swarm is an example modeling framework that works as an event-based
model. In time-driven approach, the system is updated at end of a function map. FLAME works on a time-driven approach with synchronous
updating where all agents are updated at same time and in parallel.
Asynchronous updates in a model take place when agents are updated
in random order or based on their own internal clocks.
Distributed. Multi-agent systems are concerned with distributed and coordinated problem solving. Bond and Gasser [26] describe distributed AI
in three areas:
• Distributed problem solving (DPS): how a problem is divided
among a number of nodes to be solved in parallel using knowledge
sharing.
• Multi-agent systems (MAS): concerned with ‘coordinating intelligent behavior among a collection of autonomous intelligent agents’.
• Parallel AI (PAI): concerned with performance like different computational speeds and finding new paths for problem solving.
Some agent-based modeling frameworks use CNET protocol, which work
on principle of a manager managing a set of workers. Every task is
decomposed into smaller subtasks and suitable nodes are selected to
work on the sub-task. At the end, the results are then integrated together
for a complete solution.
Decentralized behavior. Complex systems are decentralized and individuals make decisions based on their locations. Each agent evolves depending on information received locally. Whereas, evolution is based
on private memory and messages. Over time niches form, where some
agents do better than others.
Messaging. This is an important aspect of agent-based models allowing communication between agents. These interactions are responsible for emergent behavior. This follows the distributed nature of agent-based models, where messaging ensures all messages are read before decisions are
made.
Parallelism in agents. Some agent production systems use if-then statements to update rules. These rules determine the next state moved to.
A knowledge database is plugged into resolve and execute the rules.
Various parallel AI languages, like Prolog, can be used to code these
examples. However, it is important to parallelize work and synchronize
among all agents to share information. Example factors considered with
parallelism are
• Task parallelism
• Match parallelism