The management of the maintenance interventions planning is complicated. A large set of constraints must be considered when managing the plan such as the adequacy between skills and intervention acts, the respect of contracts and the availability of technicians, etc. As a result, the production of a plan becomes a big challenge, as we often have to deal with many hazards.
To address this issue, we investigate in this work the problem of planning optimization of material and human resources for maintenance interventions, and we study the capabilities of ant colony optimization algorithms to solve these problems. This approach is based on a extension of classical ant colony optimization algorithms with a local search and we introduce the notion of penalties to make forget the bad constructions of planning.
We chose to represent the problem as a Constraint Satisfaction Problem (CSP). Indeed, it allows us to reuse many tools, approaches and theoretical results about resolution and computational complexity of the problem instances. Moreover, CSP is well adapted to manage the constraints separately from the problem resolution which allows to easily customize or add constraints.
We also propose a flexible and expressive language to represent the constraints in the form of predicates that can include variables, constants and functions of the problem. All constraints encountered in the context of maintenance interventions are expressible in our system and each establishment can adapt the system to his context and add his own constraints.
The architecture of the application is modular, the definitions of the problem, of the constraints and of the resolution algorithm are decoupled. The software architecture makes it also easy to integrate another resolution algorithm (i.e. other than ACO, such as genetic algorithm for instance), we can also change the representation language of constraints to cover a more expressive or less expressive language for better performance. This allows us to easily adapt our implementation to other applications and other problems.