The complexity of distributed computing systems has raised two unprecedented challenges for system management. First, various customers need to be assured by meeting their required service-level agreements such as response time and throughput. Second, system power consumption must be controlled in order to avoid system failures caused by power capacity overload or system overheating due to increasingly high server density. However, most existing work, unfortunately, either relies on open-loop estimations based on off-line profiled system models, or evolves in a more ad hoc fashion, which requires exhaustive iterations of tuning and testing, or oversimplifies the problem by ignoring the coupling between different system characteristics (ie, response time and throughput, power consumption of different servers). As a result, the majority of previous work lacks rigorous guarantees on the performance and power consumption for computing systems, and may result in degraded overall system performance. In this thesis, we extensively study adaptive performance/power management and power-efficient performance management for distributed computing systems such as information dissemination systems, power grid management systems, and data centers, by proposing Multiple-Input-Multiple-Output (MIMO) control and hierarchical designs based on feedback control theory. For adaptive performance management, we design an integrated solution that controls both the average response time and CPU utilization in information dissemination systems to achieve bounded response time for high-priority information and maximized system throughput in an example information dissemination system. In addition, we design a hierarchical control solution to guarantee the deadlines of real-time tasks in power grid computing by grouping them based on their characteristics, respectively. For adaptive power management, we design MIMO optimal control solutions for power control at the cluster and server level and a hierarchical solution for large-scale data centers. Our MIMO control design can capture the coupling among different system characteristics, while our hierarchical design can coordinate controllers at different levels. For power-efficient performance management, we discuss a two-layer coordinated management solution for virtualized data centers. Experimental results in both physical testbeds and simulations demonstrate that all the solutions outperform state-of-the-art management schemes by significantly improving overall system performance.

The emergence of multidrug resistant bacteria world wide is a serious problem, and very few new drugs are under development. The selection of resistant bacteria is affected by factors such as mutation rate, biological fitness cost and the rate of fitness compensation. This thesis is focused on how mutation rate affects resistance to fluoroquinolones and on exploring a dosing strategy that might slow resistance development. In a set of urinary tract Escherichia coli isolates MIC values above the breakpoint for the fluoroquinolones norfloxacin and ciprofloxacin carried at least three resistance-associated mutations. In these isolates the number of resistance mutations correlated with the mutation rate. During step-wise selection for decreased susceptibility to fluoroquinolones, the accumulation of mutations in E. coli was associated with an increasing biological cost both in vitro and in vivo. However, in some lineages an additional selection step for resistance was associated with a partial restoration of fitness. During step-wise selections we found, as expected, that reduced ciprofloxacin susceptibility frequently hitchhiked with a strong mutator phenotype. More surprisingly, we also found that reduced susceptibility was frequently associated with the emergence of rifampicin-resistant populations. We hypothesise that this correlation reflects selection for fitness-compensating mutations in RNA polymerase. Mutant prevention concentration (MPC) dosing has been proposed as a strategy to reduce the selection of resistant bacterial populations. Based on limited data it had been thought that MPC might be a simple multiple of MIC, which can easily be determined. However, we showed for a collection of susceptible urinary tract E. coli that MPC could not be predicted from MIC and must be measured directly for relevant populations. Using an in vitro kinetic model we also showed that the pharmacodynamic index that best predicted prevention of resistance development in wild type E. coli was an AUC/MPC of > 22 for ciprofloxacin.

Modern air battles are very dynamic and fast, and put extreme pressure on pilots. In some unpredictable situations, like new discovered threats or mission plan deviation because of enemy aircraft, the pilots might need to replan their predefined flight route. This is very difficult, if not impossible, to do since numerous factors affect it. A system that can help the pilots to do such a thing is needed. P revious work in this field has involved methods from artificial intelligence like A*-search. In this master thesis, implementation of a replanning system based on a control theory method, Model Predictive Control (MPC), is examined. Different factors influencing the path, such as terrain and threats, are included in the algorithm. The results presented in this thesis show that MPC solves the problem. As with every method there are some drawbacks and advantages, but as a summary the method is a very promising one and is worth further development. Proposals of future work and different improvements of the algorithms used here are presented in this report as well.

There are occasions when it is preferable that an aircraft flies asclose to the ground as possible. It is difficult for a pilot to predict the topography when he cannot see beyond the next hill, and this makes it hard for him to find the optimal flight trajectory. With the help of a terrain database in the aircraft, the forthcoming topography can be found in advance and a flight trajectory can be calculated in real-time. The main goal is to find an optimal control sequence to be used by the autopilot. The optimization algorithm, which is created for finding the optimal control sequence, has to be run often and therefore, it has to be fast. This thesis presents a terrain following algorithm based on Model Predictive Control which is a promising and robust way of solving the optimization problem. By using trajectory optimization, a trajectory which follows the terrain very good is found for the non-linear model of the aircraft.

The purpose of this thesis is to investigate an alternative control strategy for a multi-variate non-linear process in a paper machine called the headbox. The proposed solution was intended to be able to be adopted on two different headbox types, currently controlled by different concepts. The methodology was to first create black-box models of the two different systems based on measurements, at one working point. Secondly, various control strategies were investigated. A more sophisticated multi-input multi-output controller MPC, or model predictive control, and a less sophisticated one, a single-input single-output, decentralised PI-controller. With help of simulations the performances of the both strategies were tested. Finally, only the decentralised control solution was implemented and evaluated through trial runs on a pilot machine. The main issue regarding the decentralised controller was the input-ouput pairing. Since the multi-variate system had four outputs and only three inputs, analysis had to be made in order to select three of those four, to form a square system. This analysis was based on the relative gain array (RGA). The resulting performance of the decentralised controller showed stability and adequate response times, surpassing the older system and making one component obsolete through the pairing changes. The MPC controller showed even better performance during simulations and shall also be taken into account if further investigatin is possible.

This thesis discuss the possibility to position control a rotor levitated with active magnetic bearings. The controller type considered is model predictive control which is an online strategy that solves an optimization problem in every sample, making the model predictive controller computation-intense. Since the sampling time must be short to capture the dynamics of the rotor, very little time is left for the controller to perform the optimization. Different quadratic programming strategies are investigated to see if the problem can be solved in realtime. Additionally, the impact of the choices of prediction horizon, control horizon and terminal cost is discussed. Simulations showing the characteristics of these choises are made and the result is shown. / Det här examensarbetet diskuterar möjligheten att positionsreglera en rotor som leviteras på aktiva magnetlager. Reglerstrategin som används är modellbaserad prediktionsreglering vilket är en online-metod där ett optimeringsproblem löses i varje sampel. Detta gör att regulatorn blir mycket beräkningskrävande. Samplingstiden för systemet är mycket kort för att fånga dynamiken hos rotorn. Det betyder att regulatorn inte ges mycket tid att lösa optimeringsproblemet. Olika metoder för att lösa QP-problem betraktas för att se om det är möjligt att köra regulatorn i realtid. Dessutom diskuteras hur valet av prediktionshorisont, reglerhorisont och straff på sluttillståndet påverkar regleringen. Simuleringar som visar karakteristiken av dessa val har utförts.

This master’s thesis describes the development of a model of the catalystsystem aiming at control by an MPC. A well functioning model, which is suitable in control purpose, is important while emission legislation become more and more hard to fulfill for the car manufacturers. Much research has been done in the field of physical modeling of the system, but in this work a linear adaptive time discrete ARX-model is developed and validated. The systems tendency to change its dynamic during usage implies that the model must be adaptive. The developed model proved to be well functioning and shows promising conditions for the MPC design. The system and the model are analyzed in the time- and frequency domains and the model is both implemented and validated in a Saab 9-5. The work has been performed both at Saab Automobile Powertrain AB in Södertälje and in Vehicular Systems Dept. of Electrical Engineering at Linköpings University.

This thesis studies the negative impact that control signal saturation may have on a controlled system. Different methods that are used to compensate for this problem are also studied and evaluated. Both sensitivity to disturbances and the effect the method has on the systems'ability to follow a reference signal will be examined. Stability will be discussed, but no conclusions whether the systems are stabilized or not can be drawn. Control signal saturation will lead to a slower behavior in general. For controllers with integral action this performance degradation will cause an extended growth in the integrating part of the controller. This leads to large overshoots and possibly to instability. As an alternative to the more ad-hoc based methods, model based predictive control is studied. This metod can explicitly handle constrained control signals. Here, too, sensitivity to disturbances and the effect the method has on the systems'ability to follow a reference signal is examined.

The situation when driving in dense traffic and at low speeds is called Stop and Go. A controller for automatic following of the car in front could under these conditions reduce the driver's workload and keep a safety distance to the preceding vehicle through different choices of gear and engine torque. The aim of this thesis is to develop such a controller, with an additional focus on lowering the fuel consumption. With help of GPS, 3D-maps and sensors information about the slope of the road and the preceding vehicle can be obtained. Using this information the controller is able to predict future possible control actions and an optimization algorithm can then find the best inputs with respect to some criteria. The control method used is Model Predictive Control (MPC) and as the name indicate a model of the control object is required for the prediction. To find the optimal sequence of inputs, the optimization method Dynamic Programming choose the one which lead to the lowest fuel consumption and satisfactory following. Simulations have been made using a reference trajectory which was measured in a real traffic jam. The simulations show that it is possible to follow the preceding vehicle in a good way and at the same time reduce the fuel consumption with approximately 3 %.

The purpose with this masters thesis was to model the scanner in a system for laser bathymetry. The model was then used to develop a controller for the scanner so a good search pattern was achieved. Two different types of models have been tested, a physical model and a Black Box model of Box Jenkins type. The physical model has been derived from Lagranges equations. Identification experiments have been used to compute the Black Box model and to find the unknown parameters in the physical model. Three different controllers have been tested, a PID controller, a model predictive controller and a controller with feedforward. The controller with feedforward gave the best result. By softening the reference signal and using feedforward a good search pattern was achieved.