Contention scheduling (Cooper & Shallice 2000) is a well-known action selection mechanism in cognitive science. It can account for normal action sequencing of daily routine actions in humans as well as for various errors exhibited by impaired human subjects. In this paper, we examine the potential of contention scheduling as an action selection mechanism for artificial agents, in particular robots. We first introduce the APOC architecture framework in order to summarize-in an "architecture-neutral" specification-the basic properties of the contention scheduling model. Then we analyze various aspects of contention scheduling that may cause problems in the context of the design of artificial agents and suggest modifications that may be able to overcome the difficulties, concluding that the contention scheduling model, as it stands, is not yet an appropriate candidate for action selection in artificial agents.
展开▼
