An Improved Discrete Particle Swarm Optimization for Airline Crew Rostering Problem

In this paper, an improved Discrete Particle Swarm Optimization (IDPSO) is presented to the Air Crew Rostering Problem for balancing crew cost, workload deviation and cooperation deviation. In IDPSO, a binary particle coding is adopted to generate initial particles. XOR-based updating rules used in updating velocity and position is to accelerate convergence rate. A selective neighbour search is employed at particles with poor performance to keep solutions qualified. Moreover, a refreshing mechanism is applied to overcoming the problem of particles trapped into local optimum and improving the diversity of the swarm. To evaluate IDPSO, computational tests have been performed , and the experiment results have proved its effectiveness.