Flexible Latching: A Biologically-Inspired Mechanism for Improving the Management of Homeostatic Goals

Controlling cognitive systems like domestic robots or intelligent assistive environments requires striking an appropriate balance between responsiveness and persistence. Basic goal arbitration is an essential element of low level action selection for cognitive systems, necessarily preceding even deliberate control in the direction of attention. In natural intelligence, chemically regulated motivation systems focus an agent's behavioural attention on one problem at a time. Such simple durative decision state can improve the efficiency of artificial action selection by avoiding dithering, but taken to extremes such systems can be inefficient and produce cognitively implausible results. This article describes and demonstrates an easy-to-implement, general-purpose latching method that allows for a balance between persistence and flexibility in the presence of interruptions. This appraisal-based system facilitates automatic reassessment of the current focus of attention by existing action-selection mechanisms. The proposed mechanism, flexible latching, drastically improves efficiency in handling multiple competing goals at the cost of a surprisingly small amount of additional code (or cognitive) complexity. We discuss implications of these results for understanding natural cognitive systems.