In the process of developing a university’s weekly course timetable, changes in the data, such as the available time periods of professors or rooms, render the timetable infeasible, requiring the administrators to repair or update the timetable. Since such changes almost always occur, it would be a sensible approach to identify a robust initial timetable, that is, one that can be repaired by making a limited number of changes, while still maintaining a high solution quality. This article formulates the problem as a bi-criteria optimization one, in which robustness is a stochastic objective, and the goal is to identify a good approximation to the Pareto frontier. It is assumed that multiple data changes, or disruptions, of multiple types can occur. The solution approach is a multi-objective simulated annealing (MOSA) algorithm, where a surrogate measure is used to approximate the robustness objective. Inspired by the concept of slack in machine and project scheduling, ten alternative measures of slack and a total of thirty surrogate measures are defined. Preliminary computational experiments are used to narrow the list of promising ones first to eight and then to two measures, which are then tested within a MOSA algorithm. Computational experiments show that one of these measures, when implemented in a multi-start MOSA algorithm, consistently provides the best Pareto frontier.

在制定大学每周课程时间表的过程中，数据的变化，如教授或房间的可用时间段，导致时间表不可行，需要管理员修复或更新时间表。由于此类更改几乎总是会发生，因此确定稳健的初始时间表将是一种明智的方法，即可以通过进行有限数量的更改来修复该时间表，同时仍保持较高的解决方案质量。本文将这个问题表述为一个双准则优化问题，其中鲁棒性是一个随机目标，目标是确定一个对帕累托前沿的良好近似。假设可能发生多种类型的多种数据更改或中断。解决方法是多目标模拟退火（MOSA）算法，其中使用替代度量来近似稳健性目标。受机器和项目调度中松弛概念的启发，定义了十个替代松弛度量和总共三十个替代度量。初步计算实验用于将有希望的列表首先缩小到八个，然后缩小到两个度量，然后在 MOSA 算法中进行测试。计算实验表明，当在多起点 MOSA 算法中实施时，这些措施之一始终提供最佳的帕累托前沿。初步计算实验用于将有希望的列表首先缩小到八个，然后缩小到两个度量，然后在 MOSA 算法中进行测试。计算实验表明，当在多起点 MOSA 算法中实施时，这些措施之一始终提供最佳的帕累托前沿。初步计算实验用于将有希望的列表首先缩小到八个，然后缩小到两个度量，然后在 MOSA 算法中进行测试。计算实验表明，当在多起点 MOSA 算法中实施时，这些措施之一始终提供最佳的帕累托前沿。