Real-time guarantees in high-level agent programming languages
Publication Date
July 13, 2011
Creator
Abstract
In the thesis we present a new approach to providing soft real-time guarantees for Belief-Desire-Intention (BDI) agents. We analyse real-time guarantees for BDI agents and show how these can be achieved within a generic BDI programming framework.
As an illustration of our approach, we develop a new agent architecture, called AgentSpeak(RT), and its associated programming language, which allows the development of real-time BDI agents. AgentSpeak(RT) extends AgentSpeak(L) [28] intentions with deadlines which specify the time by which the agent should respond to an event, and priorities which specify the relative importance of responding to a particular event. The AgentSpeak(RT) interpreter commits to a priority-maximal set of intentions: a set of intentions that is maximally feasible while preferring higher priority intentions. Real-time tasks can be freely mixed with tasks for which no deadline and/or priority has been specified, and if no deadlines and priorities are specified, the behavior of the agent defaults to that of a non real-time BDI agent. We perform a detailed case study of the use of AgentSpeak(RT) to demonstrate its advantages. This case study involves the development of an intelligent control system for a simple model of a nuclear power plant. We also prove some properties of the AgentSpeak(RT) architecture such as guaranteed reactivity delay of the AgentSpeak(RT) interpreter and probabilistic guarantees of successful execution of intentions by their deadlines.
We extend the AgentSpeak(RT) architecture to allow the parallel execution of intentions. We present a multitasking approach to the parallel execution of intentions in the AgentSpeak(RT) architecture. We demonstrate advantages of parallel execution of intentions in AgentSpeak(RT) by showing how it improves behaviour of the intelligent control system for the nuclear power plant. We prove real-time guarantees of the extended AgentSpeak(RT) architecture.
We present a characterisation of real-time task environments for an agent, and describe how it relates to AgentSpeak(RT) execution time profiles for a plan and an action. We also show a relationship between the estimated execution time of a plan in a particular environment and the syntactic complexity of an agent program.
Item Type
ethesis
Thesis Type
PhD
Supervisors
Subjects (LC)
Associated Schools / Departments
School of Computer Science (UK)
eprints ID
13036
UoN Repository URI
Except where otherwise noted, this item's license is described as
File(s)![Thumbnail Image]()
Name
546491.pdf
Type
Full-text
Size
6.86 MB
Format
Adobe PDF
Checksum ()