Agent based systems integrated with distributed computing are becoming
an essential part of AI
programming, and promise to replace complex systems such as monitoring
systems, continuous diagnostic systems, process control systems, decision
support systems, and Internet based search engines. A group of agents
can be mapped on a one or more computers to perform very complex
task.
In such a system, there are many major issues: providing fault tolerance to provide reliability in face of computer failure, development of user friendly languages, handling of multi-agent systems and programming multi-agent based systems. This course will provide the foundation for Fault Tolerant distributed computing, provide foundations for belief based multi-agent systems, review the current agent based languages and programming paradigm, and discuss research issues involved in the development of foundation and languages for fault tolerant distributed AI systems.
Introduction to Fault Tolerance: phases in fault tolerance, overview of hardware fault tolerance, reliability and availability; Distributed Systems: system model, interprocess communication, ordering of events and logical clocks; Basic Building Blocks: Byzantine agreement, synchronized clocks, stable storage, fail stop processors, failure detection and fault diagnosis, reliable message delivery; Reliable Atomic and Causal Broadcast: reliable broadcast, atomic broadcast, causal broadcast; Recovering a consistent state: Asynchronous checkpointing and roll back, distributed checkpointing, replication, semantic checkpointing.
Introduction to agents, Reactive agents, Agent Architectures, Belief systems consistency, Beliefs, desires, and intentions in agent based systems, Multi-based systems, Co-operating logical agents, behavior based agents, Agent based languages and their evaluation, agent based searches, agents and interaction, Adaptation and learning in Multiagent based systems, Fault Tolerance and recovery in multi agent based systems
Grading Assignments (30 %), One mid term (30 %), Final project (40 %)