Tasking Human Agents: A Sigmoidal Utility Maximization Approach for Target Identification in Mixed Teams of Humans and UAVs
Michael Donohue, Cédric Langbort · 2009
We propose a general utility maximization framework, based on experimentally observed human specific speed-accuracy trade-offs, to account for and exploit some characteristics of human operators engaged in human/machine mixed teams and increase their performance. In particular, we consider instances where a human operator is tasked by image capturing machines and must render a decision based upon the images. We then study methods, based on exact Karush-Kuhn-Tucker (KKT) necessary conditions and sum-of-squares (SOS) relaxations, to try and solve the resulting non-concave maximization problem, and optimally allocate a human operator’s time to identification tasks so as to maximize the probability of correct identification.