讲座时间：2025年1月14日（周二）10：00-12：00

讲座地点：创新港校区涵英楼5-6072

讲座嘉宾：Manuel Iori教授

讲座主题：A Predict-then-Optimize Approach for the Research of Underground Water on Mars

讲座内容简介：Satellite scheduling problems have been extensively studied to optimize the use of onboard instruments in space missions, with a primary emphasis on missions focused on Earth observation. In this work, we study the scheduling problem encurred by the radar MARSIS, onboard the Mars Express mission, which observes the subsurface of Mars to map the presence of underground liquid water. The resulting Mars Observation Scheduling Problem (MOSP) aims to optimally schedule MARSIS observations to reach a maximal quality coverage of the South Pole of Mars. The quality of future observations is not known in advance, but can be inferred from environmental and geographical features. To solve MOSP, we propose a predict-then-optimize framework. We first employ a machine learning algorithm to predict the quality of future observations, starting from a massive historical dataset. We then look for an optimal schedule by means of constructive heuristics, an integer linear program, and several variants of a fix-and-optimize matheuristic. The proposed algorithms are extensively tested on real-world instances from the Mars Express mission, involving two years of scheduling and an area of more than two million square kilometers, demonstrating excellent performance. In addition, several sensitivity analyses are carried out to highlight the potential for future applications.

我们研究由火星快车任务搭载的火星地下探测和电离层探测雷达（MARSIS）所产生的调度问题，该雷达对火星地下进行观测以绘制地下液态水的分布图。由此产生的火星观测调度问题（MOSP）旨在对火星地下探测和电离层探测雷达的观测进行最优调度，以实现对火星南极的最高质量覆盖。未来观测的质量事先并不可知，但可以从环境及地理特征中推断出来。为解决火星观测调度问题，我们提出了一个 “先预测后优化” 的框架。我们首先运用一种机器学习算法，从大量历史数据集中着手预测未来观测的质量。然后，我们借助构造启发式算法、整数线性规划以及固定与优化混合启发式算法的若干变体来寻找最优调度方案。所提出的算法在火星快车任务的实际案例中进行了大量测试，涉及为期两年的调度以及两百多万平方公里的区域，展现出了卓越的性能。此外，还开展了若干敏感性分析，以凸显其在未来应用中的潜力。

讲座嘉宾简介：Prof.Manuel Iori has worked in the field of Operations Research at the DISMI-UNIMORE since 2006. He joined DISMI after graduating in Management Engineering at the University of Bologna. His research activity lies in the field of Operations Research and concerns the development of decision support systems for logistics and production, with a particular focus on optimization methods and their integration with data-science techniques, machine-learning and forecasting techniques. He published more than 90 papers in international journals, including Operations Research, INFORMS Journal on Computing, Transportation Science, European Journal of Operational Research, and Computers & Operations Research. Bibliometric indices: h-index 39 and 5540 citations on Google Scholar; 121 articles, h-index 29.

欢迎广大师生积极参加！