Home Health Care Stochastic Routing Problem In Dense Communities

As the aging problem worsens in China,home health care institutes are facing increasing demands from quantitative and qualitative aspects,therefore this paper focused on a novel yet highly complex problem in such field: the Home Health Care Stochastic Scheduling and Routing Problem(HHSSCRP)in Dense Communities.This paper firstly aims to maximize the fulfilled demands with caregivers' waiting time being constrained for the reason that such problem bears more waiting costs than travel costs in communities with intensive population.Secondly,in order to raise the elders' sense of satisfaction,this paper proposes matching rules of preference and acquaintanceship upon the basic matching rule of skill-demand levels.Thirdly,the service time is assumed as normal distribution so as to better formulate the real-world environment.Since the SHHCRP is essentially a Stochastic Sequential Decision Problem,this paper employed an embedded model combined by markov decision processes and chance constrained programming.First off,the interior model is a chance constrained programming model that defines the objective that is to maximize the fulfilled demands and two crucial constraints that probabilistically limit the waiting time and workload for individual routing.And then the exterior model is a MDP model in which a caregiver is chosen as an action at each state,a reward is obtained according to the amount of demands executed by the caregiver,and the terminal is the state being either out of demands or out of caregivers.The algorithm designed for the model consists of an Ant-Colony-Optimization procedure that solves the chance constrained model and a Q-learning procedure that optimizes the policy of the MDPs.In experiments,parameters of the algorithm are analyzed and then fixed on a relatively appropriate setting that supports the following experiments of model parameters,large-scale instances and rationality analysis.At last,this paper provides precise managerial implications,suggesting that managers should: 1)attach importance to waiting costs,2)using control experiments in scheduling,3)attach importance to human factors and 4)attach importance to collecting and managing data.