This paper describes the experimental DAI development system MACE (Multi-Agent Computing Environment) using the implementation of a distributed blackboard system as an example. MACE is an instrumented testbed for building DAI systems at different levels of granularity. MACE agents run in parallel, and communicate via messages. They have have facilities for knowledge representation (e.g., models of other agents) and reasoning. The MACE environment maps agents to processors, handles inter-agent communication, and provides a language for describing agents, tracing and instrumentation, a facility for remote demons, and a collection of system-agents which construct user-agents from descriptions, monitor execution, handle errors, and interface to a user. MACE is implemented on a 16-node INTEL SYM-1 large-memory hypercube and in a Lisp machine environment. We have used MACE to model lower-level parallelism (several distributed production rule systems) and to build higher-level distributed problem-solving architectures (distributed blackboard and contract-net schemes).
[1]
Daniel D. Corkill,et al.
GBB: A Generic Blackboard Development System
,
1986,
AAAI.
[2]
Victor R. Lesser,et al.
The Hearsay-II Speech-Understanding System: Integrating Knowledge to Resolve Uncertainty
,
1980,
CSUR.
[3]
Victor R. Lesser,et al.
The Distributed Vehicle Monitoring Testbed: A Tool for Investigating Distributed Problem Solving Networks
,
1983,
AI Mag..
[4]
Barbara Hayes-Roth,et al.
A Blackboard Architecture for Control
,
1985,
Artif. Intell..
[5]
Les Gasser,et al.
MACE: A Flexible Testbed for Distributed AI Research
,
1987
.