Data Driven Automated Scheduling under Correlated Uncertainty

该项目涉及相关不确定性下数据驱动的自动化调度的设计和开发，属于EPSRC数学科学领域的统计与应用概率、数学分析和人工智能。相关不确定性在实践中经常出现，例如不确定交通下的路由、依赖天气的调度、传感器放置和污染测量、社会网络中的扩散等。这个问题吸引了人们对数据驱动方法和自动计划与调度（APS）方法进步的兴趣。APS是决策理论中的一项关键技术，在该技术中，解决方案必须在多维空间上进行优化，以综合预测的计划或行为，以实现某些预期目标。APS中的典型方法要么对导致保守决策的数据视而不见，要么做出在样本外表现不佳的分布假设。数据驱动的方法提供了一种中间立场，它涉及基于部分可用信息对不确定输入参数的估计。这可以与数据中的相关性相结合，以构建调度模型，从而实现健壮的决策，而不会变得过于保守。主管在APS和Optimisation方面有互补的经验，非常适合监督这个项目。

The project involves the design and development of data-driven automated scheduling under correlated uncertainty, and falls within the EPSRC Mathematical Sciences areas of Statistics and Applied Probability, Mathematical Analysis, and AI.Correlated uncertainty frequently arises in practice, such as routing under uncertain traffic, weather dependent scheduling, sensor placement and measurement of pollution, diffusion in social networks, among others.The problem attracts interests in methodological advances for data-driven approaches and automated planning and scheduling (APS). APS is a key technique in decision theory, in which a solution must optimise over a multidimensional space to synthesize a schedule or behaviour that is predicted to address some desired goals.Typical approaches in APS are either oblivious to data leading to conservative decisions, or make distributional assumptions that perform poorly out-of-sample. Data-driven approaches offer a middle ground that involve an estimation of uncertain input parameters based on partially available information. This can be coupled with correlations within the data to build scheduling models that enable robust decision-making without becoming too conservative.The supervisors have complementary experience in APS and Optimisation, and are perfectly suited to supervise this project.