An optimization and simulation method for the home service assignment, routing, and scheduling problem with stochastic travel time, service time, and cancellation

TL;DR

This paper introduces an integrated optimization and simulation framework for efficiently solving the complex home service assignment, routing, and scheduling problem under uncertainty, improving decision-making in service operations.

Contribution

It presents a novel two-stage solution combining column generation and simulation-driven appointment scheduling for the H-SARA problem, addressing stochastic uncertainties.

Findings

Effective column generation algorithm with high-quality initial solutions.

Simulation approach achieves desired on-time arrival probabilities.

Flexible decision support tool implemented in AIMMS.

Abstract

In the Home Service Assignment, Routing, and Appointment scheduling (H-SARA) problem, a set of homogeneous service teams must visit a set of customers. The home service provider needs to decide how many teams to hire (i.e., sizing problem), how to assign service teams to customers (i.e., assignment problem), how to route service teams (i.e., vehicle routing problem),and how to schedule the appointment times for the customers (i.e., appointment scheduling problem) such that the total cost is minimized. To tackle the H-SARA problem, we propose an efficient solution method that comprises two stages. In the first stage, we present a column generation algorithm to solve the sizing, assignment, and routing problem. The algorithm is enhanced by a high-quality initial solution which is found using the route-first cluster-second principle and a polynomial-time 2-approximation algorithm. In the second stage, due to the stochastic nature of travel time, service time, and cancellation, we propose a simulation-driven approach to decide the appointment times such that a desired on-time arrival probability is achieved. To ensure the suitability of the simulation model, we discuss the characterization of the stochastic parameters. The proposed ideas can be embedded in different solution schemes, including a fast heuristic method that finds good solutions within seconds or a more elaborate algorithm to find near-optimal solutions at the expense of longer computational time. At last, we provide a high-level flexible decision support tool implemented in AIMMS.