Adaptive Intelligent System Satisficing Cycles

Adaptive Intelligent System Satisficing Cycles

The satisficing reasoning cycle was developed to provide the guaranteed maximum latencies required for real-time performance. It is an expansion of the Dynamic Control Architecture. The Agenda Manager, instead of always exhaustively identifying all possible operations on each cycle, may use best-first search to consider as many as possible until a dynamic interrupt conditions occur. The scheduler, instead of always choosing the optimal operation from a complete agenda, may take the best it can from an incomplete one. The executor, instead of executing all actions to completion, may partially execute the next action until any of its dynamic interrupt conditions occur. When interrupted, it saves the state of the operation so that it can be continued later. The rate and importance of interrupts is determined by a set of cycle parameters which are determined by the reasoning operations. (This is a form of meta-reasoning.) By appropiately setting these parameters, the reasoning cycle can modify the effectivenss of the satisficing cycle by utilizing different reasoning policies such as:

However, the satisficing cycle does make sacrifices for its improved response times. The resulting actions are not necessarily rational with respect to the system's current control plan. The AIS developers are clearly willing to give up optimality in return for time-stress responsivity, graceful degradation, and real-time execution.
Return to the top of this architecture.

Current Location: Adaptive Intelligent Systems - Architecture - Satisficing Cycles

Go to NEXT page.