Because of this, the modules can also decrease efficiency if they must operate sequentially. Although the basic agent's modules operate in parallel, there may be situations in which one module cannot proceed until another completes a computation. For instance, the planner may require text input to be processed to extract relevant information. A completely integrated agent could aim its interpretation computations toward the planner's needs, acquiring the important data more rapidly. This is clearly a complex task and would seriously complicate the basic agent's design, though it is clearly an issue for future research.
Return to the top of this architecture.
Go to a Discussion of this issue for multiple architectures.