Dynamic re-scheduling for home health care vehicle routing problem
Résumé
Home health care services are growing in importance and number. They serve the double objective to reduce hospitalization costs, as well as improving patients' well-being. However, they face an important problem with admissions. They are often required to give a definitive approval within 24 hours, which includes paying the patient's domicile a visit in this short delay, as well as scheduling several other visits for treatments, nursing and coordination. Scheduling a crew of nurses for a week of visits, including travel times and time windows constraints, constitutes a so called Vehicle Routing Problem, and some methods exist to tackle it. However, an admission, due to its unforeseeable nature, cannot be planned in advance. So, for each new admission, we need to re-schedule in real-time, to include the new visits, without altering the existing schedule too much. In this work, we try to solve this re-scheduling for VRP by using an Ant Colony Optimization, which had already been successfully used in VRP with similar constraints. The needed proximity with the previous planning can be secured by using this previous planning to initiate the pheromones levels and by including this constraint in the quality score. Good use of the different parameters also allows using the same algorithm for the different time horizons: on-line scheduling for the admission visit, with low tolerance on planning alteration or more time-consuming in-depth research for posterior visits. So, in any case, this approach allows us to have good results in various situations.