Using Ant Colony Optimisation for infrastructure maintenance scheduling

For the optimal planning of maintenance schedules for infrastructural buildings (bridges, tunnels, etc.) in urban road systems it is necessary to take not only the budget has to be considered but also the impact on traffic flow into account to avoid unnecessary traffic jams. In an ongoing research project we develop an optimisation tool for this multi-objective problem based on Ant Colony Optimisation. In each iteration, the ants produce several different maintenance schedules for the maintenance over the next few years. Each of these schedules is formed by several scenarios of simultaneously closed roads. Parallel maintenance on different buildings can be modeled be introducing teams of ants. The scenarios are evaluated by an external traffic simulator. The quality of the different schedules, assessed by the waiting time created in the system, influences the amount of pheromone deposited on each schedule and accordingly the probability that this or a similar schedule is chosen by the ants in the next iteration step. The building condition also has influence on the probability of choosing a certain schedule: buildings in bad condition are getting more attractive to be chosen-thus avoiding that only buildings in good condition and therefore with low repair costs are scheduled for maintenance while buildings in bad condition are left to further deterioration. Additional constraints, e.g. budget constraints, can be introduced by applying a modification of the Elitist Ant strategy that guides the ants away from infeasible schedules.