The course addresses the following problem: How to estimate the dynamic quantities in a physical process given the data from a sensory system? Although the applications are wide (ranging from production processes, water management, orbit determination, telecommunication and so on), the course will concentrate on robotic applications: navigation and tracking. Especially, the SLAM problem will be addressed. SLAM = simultaneous localisation and mapping, e.g. a mobile robot that has to navigate within an unseen environment. The course will familiarise the student with methods for the estimation of state variables in dynamic systems. The course starts with an introduction of the topic 'parameter estimation' which is the fundament for state estimation. After that, the estimation paradigm will be embedded in a dynamic framework. For linear-Gaussian systems this leads to the well-known Kalman filter which is an online estimation method. An extension of the Kalman filter makes it applicable to offline estimation, and to prediction. For nonlinear dynamic systems, the so-called 'extended Kalman filter' is a suboptimal solution which only works well if the nonlinearities are not severe and the disturbances are near Gaussian. Another estimation method is the 'particle filter'. This method is generally applicable, and is optimal, but it is computationally more intensive. An important aspect of the course is bringing a theoretical concept to a practical solution. Students that attend this course will design an estimator for a given navigation process. Various estimation methods (e.g. Kalman, extended Kalman, particle filtering) will be tested and evaluated with a tracking problem and SLAM problem. Matlab is used as a development platform.

Contents
Estimation, Kalman filter, extended Kalman filter, Particle filter, prediction, SLAM.