Dynamic resource allocation in human queueing systems through market-based control

Many complex systems have both social and technical work system aspects which impact system performance. However, much prior work focuses on only one of these two aspects while largely neglecting the other. This work seeks to develop and demonstrate a holistic method of dynamic resource allocation in systems having queueing dynamics using the important domain of debris removal following natural disasters. This system requires dynamic resource allocation in the sense that the system state (i.e., travel times and service rates) evolves over time, thereby requiring resource adjustments to maintain system efficiency. The primary social consideration in the case of dynamic resource allocation is the performance effects of varying team composition (e.g., team size and working history of members). The technical work system itself (a queueing network) also has dynamics which significantly influence the performance of an allocation strategy. Furthermore, the entities flowing in this system are teams of people within an organizational structure. In many applications, these actors within the system are directed by dispatchers who each exercise considerable autonomy. System control is therefore achieved indirectly through designed market incentives (i.e., market-based control).;This dissertation draws on and extends two bodies of literature (team performance and routing in queueing systems) in order to develop market-based control policies (reward structure) for these complex queueing systems. Human decision makers are not likely to reliably make rational, or optimal, decisions with respect to the reward structure. Thus, computational models of the decision makers, along with models of the team composition performance effects and the queueing network, are created which enable a simulation study of the effects of differing levels of rationality. The work consists of both empirical and theoretical research in which the theoretically optimal market incentives will be updated in light of the resulting team compositions. The work is organized into three research activities: 1) a field study of team composition performance effects --- the social aspect of the system; 2) analytic work in optimal allocation of entities in closed queueing networks which, assuming rational decision makers, provides insight into optimal market incentive structures --- the technical work system aspect of the system; and 3) a simulation study which integrates these two aspects of the system and explores the effect of bounded rationality on the part of human decision makers as they respond to different market incentives for a variety of conditions. The models developed for the simulation study are validated with field data and thereby provide a potential mechanism for the team composition effects observed in the field study.;The contributions of this work include greater clarity of team composition performance effects by considering performance as a multi-dimensional construct, novel methods of optimal probabilistic routing and an approximation of optimal vertex (multi-class, deterministic) routing in closed queueing networks, and an exploration of the effects of bounded rationality in network games.