An Adaptive Ant Colony System for Public Bicycle Scheduling Problem

Public bicycle scheduling problem (PBSP) is a kind of problem that how to design a reasonable transportation route in order to reduce cost or improve user satisfaction under certain constraints. PBSP can be regarded as a specific combinatorial optimization problem that ant colony system (ACS) can solve. However, the performance of conventional ACS is sensitive to its parameters. If the parameters are not properly set, ACS may have the disadvantage of being easy to fall into local optimum, resulting in poor accuracy and poor robustness. This paper proposes an adaptive ACS (AACS) to efficiently solve PBSP. Instead of fixed parameters in ACS, each ant is configured with own different parameters automatically to construct solutions in AACS. In each generation, AACS regards the parameters in the well-performed ants as good parameters and spreads these parameters among the ant colony via selection, crossover, and mutation operators like in genetic algorithm (GA). This way, the key parameters of ACS can be evolved into a more suitable set to solve PBSP. We applied AACS to solve PBSP and compared AACS with conventional ACS and greedy algorithm. The results show that AACS will improve the accuracy of the solution and achieve better robustness.