Multi-period dynamic technician routing and scheduling problems with experience-based service times and stochastic customers

While home services is a fast growing industry, little attention has been given to the management of its workforce. To maintain growth, a key challenge for home-service companies is managing their expensive and limited labor resources. In particular, the time an employee needs to provide high quality service often depends on his/her experience. Importantly, experience increases over time, thus gradually decreasing the time required to provide service. By accounting for employee experience and the accompanying learning, managers can take advantage of capacity increases that result from experience, improving efficiency and enabling further growth.

We address the technician routing and scheduling problem from three perspectives, which in turn constitute the three parts of my dissertation. First, we introduce a model of technician routing that explicitly models individualized, experience-based learning. We convert the multi-day problem into a series of daily problems and approach the daily decision making in a myopic fashion. The results demonstrate that explicit modeling and the resulting ability to capture changes in productivity over time due to learning lead to significantly better and different solutions than those found when learning and workforce heterogeneity is ignored. We show that these differences result from the levels of specialization that occur in the workforce.

In the second part, we design solution methods that account for the fact that serving today's demand has implications, in terms of learning, for serving tomorrow's demand. We integrate the future information in the decision process to overcome the drawback of the myopic algorithm. We introduce the multi-period technician scheduling problem with experience-based service times and stochastic customers. Then, we model the problem as a Markov decision process and introduce an approximate dynamic programming-based solution approach. The model can be adapted to handle cases of worker attrition and new task types. Using an extensive computational study, we demonstrate the value of our approach versus a myopic solution approach that views the problem as a single-period problem.

In the final part, we continue exploring the value of integrating future information into the current period decision-making process for the Multi-period Dynamic Technician Scheduling Problems with Experience-based Service Times and Stochastic Customers discussed in Chapter 3. We propose an alternate approximate dynamic programming solution approach with basis function to approximate the value function by taking the advantage of the future information for the whole planning horizon. We turn to an offline simulation procedure to recursively update the coefficient vector of the basis function, which allows fast decision making within the execution phase. Our computational results demonstrate the value of the ADP solution approach with the basis function.