Declarative Abstractions for Agent Based Hybrid Control Systems

Abstract

Abstract. Modern control systems are limited in their ability to react flexibly and autonomously to changing situations by the complexity inherent in analysing environments where many variables are present. We aim to use an agent approach to help alleviate this problem and are particularly interested in the control of satellite systems using BDI agent programming as pioneered by the PRS. Such systems need to generate discrete abstractions from continuous data and then use these abstractions in rational decision making. This paper provides an architecture and interaction semantics for an abstraction engine to interact with a hybrid BDI-based control system.

Handling Agent Perception in Heterogeneous Distributed Systems: A Policy-Based Approach

“…In the latter case, we assume that there is a single channel or that the channels have been designed so that the buffer does not need to synchronise state updates from different channels. We also assume that all percepts in the state update are to be processed by the same policy 5 , or that it does not matter if a single state update results in different policies' outputs being perceived by the agent at different times 6 . For each percept-processing policy in use, a percept buffer maintains a (thread-safe) queue of incoming percept sets that have been pushed on the queue by the channels.…”

“…The first two issues above have been repeatedly encountered by researchers [6,8,12,13,10]. However, as yet, agent development tools do not provide any platformlevel solution to these problems, leaving the agent programmer to implement their own application-level solutions.…”

“…However, BDI agents have proven their value beyond simple simulated systems. They have been used for implementing robots [16,15], "intelligent virtual agents" [13,7,2] that control avatars in virtual worlds and multiplayer games, and even for real-time control of satellites [6]. As well as these situations where an agent's 'body' is controlled by software external to the agent, it may also be the case that an agent is only a component of a larger distributed system involving multiple technologies and protocols.…”

“…In such a case, this continuous data stream should be discretized before providing it to the agent as percepts. Such an abstraction engine has been described by Dennis et al [6] in the context of using BDI agents to control a satellite. By providing only abstract environment information and/or discretized information to an agent, the problem of cognitive overload can be minimised.…”

Agent-Based Autonomous Systems and Abstraction Engines: Theory Meets Practice

“…Translating continuous sensor data into abstractions suitable for use in BeliefsDesires-Intentions (BDI) style agent programming languages is an area of active study and research [7,6,17]. Work in [7] provides an architecture for autonomous systems which explicitly includes an abstraction engine responsible for translating continuous data into discrete agent beliefs and for reifying actions from the agent into commands for the underlying control system.…”