Jump to the MAX Overview page.
Jump to the Architectural Details Menu page.
Jump to the Home page.
Architectural Details
The architecture layer of MAX is dominated by a interpreter kernel, which is
a rule based forward-chaining system used to analyze the other architecture
components, and in so doing it uses meta-level reasoning. The kernel is
equipped with simple functions such as match, difference, and intersection
to deal with the predicates in its knowledge. Because all of the knowledge
in the system, including the problem spaces, are encoded in a uniform
declarative structure, the kernel can access and assimilate this information
into its procedural matching algorithm.
The following document outlines the structure of the MAX architecture:
The functionality of MAX primarily consists of the knowledge which implement
the capabilites of a particular MAX agent. This knowledge consists of:
behaviors which are analagous to problem spaces in Soar(sets of operators
and rules), monitors which are like behavior rules but which can be defined
dynamically and run in the background, and a storage area for current behavior
which allows for self-reflection. Other knowledge can also be supported by
a MAX agent.
Some of the capabilites that have been integrated into an architecture of this
type are:
- focus of attention
- multiple problem solving strategies
- execution monitoring
- domain refinement experimentation
- explanation based learning
- process interruption and intelligent recovery