Theo

• Homogenous knowledge representation

  All problem-solving in Theo is inference of slot values for frames, so all problems and solutions are represented in the same medium and in a similar manner. This allows the system to reason about and operate on not only its input data, but also on its control knowledge.

• Efficient knowledge access

  Theo uses a caching mechanism to store the rote knowledge that it aquires through learning in its knowledge base. This allows fast, efficient access to any of the factual knowledge that the system learns.

• Knowledge consistency

  Theo implements a simple form of truth maintenance. Dependancies for entries in the knowledge base are saved in the DEPENDANCIES slot of the frame-based representation. When the value of a piece of knowledge in the system is changed, the values of all the knowledge dependant on the changed knowledge are uncached, keeping the knowledge of the system consistent.

• Meta-reasoning

  As mentioned earlier, Theo's uniform representation support meta-reasoning. Using SE , Theo's inductive learning system, Theo can order its slot-inferring methods in an attempt to optimize the order in which they should be attempted.

• Impasse driven

  As with Soar, Theo is impasse-driven. Impasses in Theo take the form of a query of the unknown value of a slot. In this case, the impasse results in the inferring of the corresponding slot value. As in Soar, an impasse in solving one problem results in the formulation of another problem in the same representation.

• Eager/lazy sensing

  Theo has the ability to control the frequency with which it samples different aspects of the environment. Eagerly sensed features are sampled at the beginning of each sense-decide-execute loop, while lazily sensed features are only sampled when needed. This allows the agent to focus on certain aspects of the environment.