Home healthcare (HHC) services require caregivers to travel from healthcare centers to patients’ homes, making routing and scheduling a critical operational problem. Rising fuel costs and sustainability targets are pushing providers to adopt electric vehicles (EVs), yet most HHC studies assume homogeneous fleets and ignore battery dynamics, charging decisions, and mandatory caregiver breaks. This paper studies an HHC routing and scheduling problem with a mixed fleet of electric and conventional vehicles under patient time windows, mandatory lunch breaks, battery consumption and charging, daily working and distance limits, and workload fairness among caregivers. A mixed-integer linear programming (MILP) model is proposed to minimize total travel distance while penalizing workload imbalance. Because the problem is NP-hard, a three-stage solution framework is developed: an exact MILP solved with Gurobi for small instances, an Adaptive Large Neighborhood Search (ALNS) with elimination strategies for larger instances, and two enhancements—a tabu memory mechanism and a machine-learning module that guides operator selection dynamically. Computational experiments on instances with 40, 50, and 60 service nodes and varying EV/conventional ratios show that the learning-guided variant (ALNS+Tabu+ML) improves average solution quality by about 7.5% over the exact model within a fraction of its computing time and converges in fewer iterations. The results confirm that integrating learning into ALNS yields faster, higher-quality solutions for sustainable home healthcare routing.

Keywords

Home healthcare routing, Electric vehicles, Mixed fleet, Adaptive large neighborhood search, and Machine learning.