Estimation, Information Fusion and Machine Learning - Cognitive Tools for Cyber-Physical Systems

Lecturer:  Guido Dartmann, A. Charlish

Type:  Master Lecture

Credits: 4

Lecture

Course language: English

Contents

Large sensor networks and cyber-physical systems are playing an increasingly important role in a large variety of modern engineering applications. Such a cyber-physical system is controlled using interconnected computational elements that perform actions based on data acquired through a range of sensor data sources. A key challenge for such a typically highly non-linear system is the ability to estimate key parameters of interest from the noisy, imperfect sensor data sources, and fuse information obtained on commonly observed quantities. Using the information obtained, the next challenge is to learn relationships and patterns from the large volume of estimated parameter data that the system produces. These challenges can be tackled using modern techniques for es-timation, information fusion and machine learning, which serve as cognitive tools enabling the system to achieve a level of perception on its operating environment. This achieved perception acts as a basis for executing well-informed actions. The combination of action and perception facilitates a cognitive, adaptable and autonomous system that can self-optimize its efficiency and functionality in run-time. The goal of this course is to provide a comprehensive overview of rele-vant techniques for non-linear estimation, information fusion and machine learning, which can be used as cognitive tools in cyber-physical systems. Modulbeschreibung: Part 1 - Introduction . Motivation and application of cyber-physical systems Part 2 - Mathematical tools for estimation and information fusion . Review of fundamentals of statistics . (Nonlinear) parameter estimation . Multi-object estimation . Statistical models . Information fusion Part 3 - Machine learning . Review of convex optimization fundamentals . Supervised learning . Unsupervised learning . Reinforcement learning and . Consensus and distributed control