Discussion

There are a number of limits to the algorithm as it is now, the main ones are listed below.

• The language provided to describe the input-output relationship of a competence module is oversimplified. There is no way to work with abstractions, neither can variables be used.

• A network does not maintain a record of its past `search'. As such the same planning mistake can be made over and over again in the same plan, making the system loop.

• It is not yet clear how, given a specific application, one can select values for the global parameters that produce the desired action selection behavior.

In the remainder of this section we discuss the importance of these limits and sketch solutions to those that represent an important limitation. The proposed solutions do not inflict on the current philosophy and the merits it has. The implementation of these solutions will be the main concern of our future research.

• Variables
• Handling Loops
• Selecting the Parameters