ABSTRACT
The ability of members on a team to reason about each others’ capabilities and workload is important for effective teamwork. This is required for proper task allocation and load balancing, as well as many other team processes such as adaptive-ness, proactive assistance, and backing-up behavior. The present work proposes to incorporate capability reasoning into intelligent agents to produce better teamwork simulations, to work better with humans as virtual team members, and to facilitate team training. However, classical models of capabilities in computational systems and intelligent agents are inadequate for representing the more complex aspects of human performance, such as the ability to perform multiple tasks in parallel, interference among these tasks, effects of limits on attention and other cognitive resources, and the ability of humans to dynamically adjust their level of effort on tasks. In this paper, we present a formal mathematical model of capabilities that accounts for these effects. The model posits finite pools of internal resources, for which tasks compete; quality of performance depends on the amount of resources allocated. Capabilities are defined according to whether a feasible schedule can be found that allows a set of tasks to be completed within given constraints (e.g. deadlines) while not exceeding the capacity of any internal resource. An extension of the model is then proposed to incorporate multiple resources.
