Skip to content

Automata Glossary

Eric Steen edited this page May 17, 2020 · 24 revisions

In Automata, an Automaton is an Agent with abstract superpowers. Our core team scours the literature on multi-agent and autonomous decentralized systems to detect common patterns and inter-relationships amongst key concepts in order to synthesize the best set of abstractions determined to be most generally valuable to trained AI architects.

Worlds

  • A World is made up of Agents (automata) which interpret & act based on global(all-agents) & local(per-agent) Obervations (State Space Representations) from the World.
    • update process produces next state based on most recent observation and most recent action
    • Potential names of global state-spaces shared across agents: context, motif.
    • Potential names of per-agent local state-spaces: strata, motif.

Ontology

  • An Ontology contains the highest level representations designed for a World. It gives us a way to provide shared goals, beliefs, features. Ontological working memory can be used to describe inter-relations in terms of perspectives, constraints, environs, stratified classes of objects/agents and their types, objective and predictive representations

Environ

  • The representations describing the World in its Ontology may determine the initial Environs and their representations and vice versa, they are dependent and can be mixed and matched.

Agents

  • Within a World, an Agent is represented as an Automaton with access to State Space Representations, and configurable Action Selection and Action Execution, and Knowledge/Memory capabilities.

Automata

  • A set of Agents configured by a World, each carrying out actions corresponding to a set of state-space representations (environments) unique to its goals. May be coordinated through a shared memory structure (blackboard).

  • Meta-level Control

    • Processing which is Triggered each heartbeat at the [automata level] to support agent interaction, and any potential network reorganization. Meta-level control is the ability of an agent to optimize its long term performance by choosing and sequencing its deliberation and execution actions appropriately.

Automaton

  • Belief State
    • A belief state encapsulates the beliefs an agent has about its current state: that is, probability distributions for each variable within the state.
Clone this wiki locally