Ensuring navigable water levels is the main concern of operational water management of inland waterways. As a secondary objective it is in most cases desirable to minimize electrical energy costs caused by pump operations, and to maximize power generation by turbines. An effective method that optimizes energy related objectives while guaranteeing safe water levels is model predictive control (MPC). Based on the operational requirements and a simulation model of the canal, an optimal control problem is formulated and an optimized water management strategy can be calculated. Here, the simulation model is based on a coarse grid Godunov-type discretization of the Saint Venant equations, which gives a good reproduction of the system behavior. In addition to a sufficiently precise model, knowledge of the current system state is needed to accurately predict the future evolution of the system and the effect of the control input. In many cases only water level measurements are available at a few points along the canal reaches. The remaining level and flow values must be estimated. Two estimation approaches are investigated. First, a moving horizon state estimator (MHE) is implemented. Because of its structural similarity to MPC, MHE can be conveniently integrated with the control algorithm and delivers solid performance. However, posing a second optimization problem that must be solved, MHE adds considerably to the overall computation time. As an alternative, an unscented Kalman filter (UKF) for state estimation of canal systems is implemented in this work, and the performance of both estimation methods is compared. Another important effect that must be considered is given by non-measurable inflows that can have a strong influence on the system, e.g. in case of heavy rain or flooding. To account for that, the system model is augmented, and additional inflows are estimated using MHE and UKF. In simulation experiments, both methods show the ability to estimate a non-measureable inflow with and without the addition of simulated measurement noise. The two state estimation algorithms are validated in combination with the MPC in several closed-loop simulations using a high resolution hydraulic model of the Mittellandkanal-Elbehavelkanal canal system in northern Germany.

Neuromuscular electrical stimulation (NMES) is the application of a potential field across a muscle via internally or externally placed electrodes in order to produce a desired muscle contraction. When the desired muscle contractions yield functional movements of the limb, the application is termed functional electrical stimulation (FES). FES is used to prevent muscle atrophy and improve limb control in people with movement disorders such as stroke and spinal cord injuries. For example, stroke patients often have difficulty walking normally due to damage to the nervous system. An effective FES treatment protocol for such a person would involve stimulating the quadriceps muscles to extend the knee joint, replicating the motions of a walking gait. There is a desire for an automatic stimulation strategy that would cause the lower limb to move in such a way as to mimic the motions of walking, but uncertainties in muscle physiology make controller development difficult. A variety of controllers have been developed, and our work has focused on the development of a nonlinear controller coined RISE for robust integral of the sign of the error. Through a Lyapunov-based stability analysis, the RISE controller is proven to yield an asymptotic stability result despite the uncertain and nonlinear muscle model and the presence of additive bounded disturbances. An overview of NMES and the challenges involved in tracking control of a human limb, development of the RISE controller and experimental results, and a look at current and future applications are presented.

In this talk, we introduce and develop effective tools based on statistical learning theory for solving difficult control problems. The existing sample size results for system design by means of semi-infinite programming using randomized strategies are generally considered to be very conservative by the control community. The main contribution of this lecture is to demonstrate that this is not necessarily the case. Utilizing as a starting point one-sided results from statistical learning theory, we obtain new bounds on the number of required samples that are manageable for reasonable values of probabilistic confidence and accuracy.

We adjust the node and edge weightings of graphs using convex optimization to impose bounds on their Laplacian spectra. First , we derive necessary and sufficient conditions that characterize the feasibility of spectral bounds given positive node and edge weightings. Synthesizing these conditions leads naturally to algorithms that exploit convexity to achieve several eigenvalue bounds simultaneously. The algorithms we propose apply to many graph design problems as well as multi-agent systems control. Finally, we suggest efficient ways to accommodate larger graphs, and show that dual formulations lead to substantial improvement in the size of graphs that can be addressed.

In mechatronics the field of oscillation damping control for flexible structures is one of the key areas resulting in interesting new functionalities. Reasons are given by the necessity of exact adjustment of systematic evaluation of dynamic properties, adequate sensor and actuator systems, efficient and capable hardware and software implementation. It can be clearly seen that this combination is an ideally reflection of the three main areas of mechatronics: mechanical engineering, electrical engineering and computer science. In the talk at four examples of oscillation damping control respectively control of mechanical structures the idea of mechatronic design from a control perspective is given. First example is the oscillation damping control for a fire turntable ladder, second example is an automated slewing crane, third example is the oscillation damping control for a reconnaissance and disarming robot system of the German army, and the fourth example is dealing with a mechatronic approach for the control of a deformable, flexible mirror in adaptive optics.

本講演ではまず6個のタンパク質濃度波形からタンパク質ネットワークを推定する手法を説明する．つぎに推定されたタンパク質ネットワークを化学反応式に基づいて非線形微分方程式を導出し，推定で用いた6個のタンパク質濃度波形が再現されることを示す．さらに，そのロバストネスについて議論する．