A comprehensive evaluation of the effectiveness and robustness of statistical methods and artificial intelligence in control applications

人工智能有着广泛的应用。新的研究试图在各种应用中实现人工智能和一般统计模型，已经取得了许多成功。然而，在任何实现的模型设计中选择的巨大自由使优化复杂化，而这类模型的统计性质往往被认为不够稳健。尤其是在现实生活中的控制中，误差容限及其后果在系统设计中至关重要。该项目建议研究为任何一般控制应用设计有效的统计控制系统的可能方法，利用它们的优点，同时最大限度地减少它们在数据收集、健壮性和可用性方面的缺点。

Artificial intelligence has a wide variety of applications. Novel research attempts to implement AI and general statistical models in a variety of applications, already with many successes. Nevertheless, the great freedom of choice in model design for any implementation complicates optimisation, while the statistical nature of such models can often be deemed not sufficiently robust. This is especially the case in real-life control, where the margin for error and its consequences are of crucial importance in system design. This project proposes to investigate possible ways to design effective statistical control systems for any general control application, exploiting their benefits while minimising their drawbacks in data collection, robustness and usability.