Queieuing: a behavioural Approach

The present proposal is a continuation of the work performed under our current grant, and is part of a broader attempt to understand the behavioural aspects of queuing theory. Traditionally, work in the queuing area has focused on the design, running and performance of facilities, with relatively little emphasis on the customers, the users of this facility. While this has been useful in helping to understand the capacity requirements and the impact of different configurations of production and service facilities, it has in many cases also been a built-in limitation and it has limited our ability to explain the behaviour observed in many real queues. This proposal continues research that we have conducted for several years, shifting the focus to the information structure and to how customers and the service provider adapt their behaviour to changing circumstances. More specifically, we study how "intelligent" customers and service providers interact. In other words, we are moving the focus from (static) facility design to reactive customers and managers. This requires us to extend traditional queueing models in the following way: current and potential customers, as well as the facility management, must be provided with information regarding the current performance of the facility and be empowered with sufficient computational capability to enable them to react to this information. Given these elements, current customers can decide whether or not it is worthwhile using the facility, potential customers can decide whether or not to become customers, and management can decide how to adjust capacity. We explore these issues in three stages.

First, we continue our current work on using aggregated information flows as feedback to provide information to current and potential customers, and to the facility management, about the current expected waiting time. The feedback view on queuing implies that customers form a view of how long it takes on average to receive service. This average will then determine whether they choose to join the facility (potential customers) or continue using it (current customers). The time lag with which customers adjust their perceptions has a major impact on system behaviour. Model complexity is further increased by allowing the manager of the facility to form his own perception of the required capacity and adjust it accordingly, again with a time delay. In this way we obtain simultaneous adjustment - i.e. both customers and management adapt their behaviour continuously to an evolving situation.

In the second stage we use agent modelling to capture the experience of individual customers, and consider competition between several facilities using agent based models. By doing so we create individual experiences for the customers (i.e. they do not react to average perceptions but to their own experience). This approach enables us to gain a better understanding of how customers might co-evolve with the facility that serves them, something which cannot be done within the traditional queuing framework. In both stages we use nonlinear models, implying that we cannot find closed form solutions and thus resort to simulation and numerical analysis to gain understanding of model behaviour.

In the third stage we resort to an experimental approach to test the insights derived from the simulation models used in stages 1 and 2 regarding the behaviour of customers and the facility managers. We believe that the project will lead to new insights and understanding of queuing systems, and thus to a better design of these systems which we experience on a daily basis.