Image
The Renaissance group operates several residential care facilities for elderly people. It focuses on improving the quality of care and working conditions through organizational innovation and digital tools in the medico-social sector.
Technical/scientific Challenge:
Nursing homes face complex scheduling challenges due to strong temporal, human, and logistical constraints, while also needing to improve both staff working conditions and resident satisfaction. The objective of this project was to rationalize staff activity planning by developing an automatic scheduling optimization solution tailored to medico-social facilities. The solution aims to ensure a fair distribution of workload and efficient staff mobilization, while increasing resident satisfaction by respecting their preferred time slots. The project also fits into a broader initiative to transfer HPC and AI optimization skills to the socio-economic sector, demonstrating the relevance of combinatorial modeling and optimization tools in a non-industrial context.
Solution:
The work was carried out using Google’s OR-Tools library, which provides several optimization solvers suited to complex scheduling problems. Two complementary approaches were developed.
First, a constraint programming approach (CP-SAT) was used to maximize resident satisfaction by assigning caregivers to time slots while respecting duration, continuity, and exclusivity constraints. Boolean variables were defined to represent task assignments between residents, time slots, and caregivers. This approach was particularly applied to personal care tasks, such as bathing, which have flexible durations (ideal and degraded). While this model validated the feasibility of the approach, it showed limitations in resolution time for large problem instances.
Second, a Vehicle Routing Problem (VRP) formulation was introduced to account for travel times between rooms, modeling caregivers as vehicles and residents as locations to visit. This approach efficiently minimizes overall task duration but does not directly maximize satisfaction.
To overcome these limitations, a hybrid strategy was adopted, where CP-SAT results guide the VRP optimization. This combined approach provides a balanced solution between quality of service and operational efficiency and represents an original scientific contribution to healthcare scheduling.
Business impact:
The initial results demonstrate the feasibility and relevance of the proposed approach. The prototype is capable of automatically generating realistic staff schedules that respect the constraints defined by EHPAD Renaissance. It currently handles personal care tasks using a dedicated model and manages breakfast activities as specific tasks. Visualization tools, including staff and resident timetables, were manually generated to facilitate interpretation and validation of the results. From a technical standpoint, CP-SAT produces high-quality schedules at a high computational cost, while VRP offers fast resolution but requires careful constraint tuning. Their combination opens the way to a robust and production-ready hybrid solution for care planning.
Benefits:
Automated generation of realistic and constraint-compliant staff schedules
Improved workload distribution and staff mobilization
Increased resident satisfaction through respect of preferred time slots
Hybrid optimization approach balancing service quality and operational efficiency
Demonstration of AI and combinatorial optimization relevance in the medico-social sector