Resource overload problems with tardiness penalty: structural properties and solution approaches
研究了在项目超期时加入延迟惩罚的资源过载问题，通过结构性质缩小搜索空间，并提出了混合整数线性模型和遗传算法，实验验证了中小规模实例的效果。
Abstract In this paper, we consider a resource overload problem and add a tardiness penalty to the objective function when a prescribed project makespan is exceeded, which enables a trade-off between a balanced resource utilization and a project delay. For the tardiness penalty, we distinguish between a constant and variable delay cost variant. Based on the structural properties of the resource overload problem, we show that the search space of the resource overload problem with tardiness penalty can also be reduced utilizing quasistable schedules. In addition, we discuss the application of these findings to further problems, which include objectives composed of a locally concave and a concave function or a reward structure for an early project completion instead of a tardiness penalty. As solution approaches, we present mixed-integer linear model formulations as well as a novel genetic algorithm with a decoding procedure, which exploits the devised structural properties. The performance of the genetic algorithm is improved by implementing learning methods and utilizing lower bounds. Finally, we present results from experiments on small to medium sized problem instances.