This thesis developes a control allocation method based on the Model Predictive Control algorithm, to be used on a missile in flight. The resulting Model Predictive Control Allocation (MPCA) method is able to account for actuator constraints and dynamics, setting it aside from most classical methods. A new effector configuration containing two groups of actuators with different dynamic authorities is also proposed. Using this configuration, the MPCA method is compared to the classical methods Linear Programming and Redistributed Pseudoinverse in various flight scenarios, highlighting performance differences aswell as emphasizing applications of the MPCA method. It is found to be superior to the two classical methods in terms of tracking performance and total cost. Nevertheless, some restrictions and weaknesses are revealed, but countermeasures to these are proposed. The newly developed convex optmization solver CVXGEN is utilized successfully in the method evaluation. Providing solve times in milliseconds even for large problems, CVXGEN makes real-time implementations of the MPCA method feasible.

To avoid hole stability problems in an increasingly fierce drilling environment, the demand for accurate control of the pressure profile during drilling operations is rising. As standard instrumentation of drilling rigs have poor measurement of the bottom hole pressure, there is a need for estimation. However, a precise model of a drilling process is difficult to obtain, so a competent observer, using a simpler, lower order model, should be satisfactory.In this master thesis several approaches on estimation are discussed together with a suggested improvement in the annular friction model. The estimators tested are: First, the moving horizon observer, which is presented together with prior work by the author and Marcel Paasche; Second, the unscented Kalman filter, which is a new estimation candidate introduced together with regularization to compensate for the slow update and lack of availability in bottom hole pressure measurements. Last, different combinations of the two observers are proposed.All observers are tested in simulations and good performance is found for both the MHE and UKF. Parameter adaptation is found to be effective for both observers, but the UKF encounters some minor observability issues when the system is not sufficiently exciting. Different combinations of the two observers increase computational complexity, unfortunately without achieving better accuracy in estimates. The estimates are deteriorated when the alternative friction model is tested, and it is thus considered a failed attempt to improve the simple third order Kaasa model.

This thesis considers two different state estimators for an unmanned surface vehicle owned by Maritime Robotics. They will both be tested together with a dynamic positioning system developed especially for the vessel in question. The observers should utilize measured acceleration delivered from an IMU, and some kind of wave filtering should be done on the estimates. The emphasis will be put on how good the achieved state estimates are, and on to what degree of wave filtering should be done to the estimates, considering they are meant for a small vessel. Both observers were tested in a Matlab/Simulink dynamic positioning simulator. Two simulation scenarios were used in the simulations, each with different weather conditions. The first observer was originally an attitude observer which was simplified specifically for this application. It makes use of the measured body-fixed acceleration as input and has feedback from position, heading and velocity. The simulation results turned out to be less than satisfying, but the observer showed potential and could be further developed. The second observer considered was based on a model of the Viknes 830. Instead of using the measured acceleration as input, the measurements were used as feedback to improve the model estimates. This observer performed better than expected in the simulations, reducing both convergence time and controller usage. The simulation results for this observer also indicated that wave filtering could be beneficial for small vessels.

This thesis contains modeling, observability analysis and state estimator design of a tethered airfoil. A existing model is extended to include wind as a state and the wind is modeled. The observability analysis shows that the system is indeed observable. The choice of state estimator is evaluated and the UKF is found to achieve best results during simulations.

Unmanned vehicles have become more applicable due to extended research and more advanced technology through the last decades. The vehicles are versatile, hence studying ocean floor and reconnaissance and surveillance are popular areas of application. To perform unmanned operations, waypointfollowing based on guidance, navigation and control systems are used in general. Unfortunately, not every waypoint setup guarantees a feasible trajectory as regards to the vehicle dynamics. Thus, algorithms analyzing the feasibility of the paths and suggesting solutions for infeasible paths have been proposed in this master’s thesis. Four different tangent vector calculation methods have been used to develop smooth, parameterized C1-continuous curves as suitable paths between the successive waypoints. The appurtenant path curvature has been calculated and compared against the vehicle dynamics and constraints. Minimum vehicle turning radius is mapped directly to maximum trajectory curvature. When waypoint placement provides infeasible trajectories, three different algorithms of moving, adding or removing waypoints have been suggested tocope with the trajectory curvature. For pure waypoint-following schemes without path generation, adapted velocity and acceptance circle radius have been applied. The adapted reference values are functions of the angle between the neighboring waypoints, and turns out to yield more efficient operations with respect to time and overshooting errors than with constant values. Employing path generation proves that the effect of preturns before approaching waypoints in addition to regulating velocity according to path curvature is beneficial to minimize deviation from waypoints. If, however, the trajectory is infeasible even after velocity adaption, one of the three path altering algorithms may be applied to provide a feasible path.The simulations have been carried out in MATLAB/Simulink on two models; a first order Nomoto autopilot model and a more complex mathematical model of a Viknes 830 unmanned surface vehicle. Despite simulating on boat models in R2, the path planning ideas and trajectory feasibility check can be employed on various types of unmanned vehicles, and -extended to R3.

The NTNU Test Satellite project aims to launch a 10x10x20 cm nanosatellite, weighing less than 2.66 kg. The satellite, which follows the CubeSat standard, is a relatively low cost project, and explores the use of comercially available, low cost electronic components in space. This work is mainly concerned with power management and distribution for NUTS, but also describes the design of a system backplane.The thesis is split into two parts. The first describes the design and evaluation of the backplane with main focus on power distribution, and the second part describes battery charging and power conversion in the Electrical Power System (EPS) unit.The backplane has been designed with flexibility, fault-tolerance and simplicity in mind. The design is based on a single I2C bus with bus repeaters for each sub-module, with the ability to isolate individual modules from the system in case of a malfunction. Power is distributed with dual 3.3V and dual 5V buses working in active redundancy, ensuring continued operation should a voltage converter fail. Power distribution for each module consists of three parts: power supply or-ing, current-limit switch and power monitor, and is integrated into the backplane. The state of the power switches and bus repeaters are controlled from two master modules, and a watchdog timer ensures return to a default state should both master modules be disabled. The proposed solution integrates all the management logic and power distribution circuitry into the backplane itself, leaving the valuable PCB space on the sub-modules free for module implementations. The power distribution unit has been evaluated with a separate prototype, and displays good performance in terms of power losses and response to over-current conditions. A prototype implementation of the backplane has been produced.The electrical power system of the satellite consists of solar panels, batteries, and voltage converters. With very limited power available, the main focus of the design has been to reduce losses in power conversion. A battery charge regulator (BCR) has therefore been proposed, which charges the battery while keeping the solar panels at their maximum operating point. This integrates a Maximum Power Point Tracking stage with a battery charger into a single converter, greatly increasing efficiency. Redundancy can be achieved by using separate converters for separate panels in parallel. A prototype BCR has been tested, and initial results show over 91% converter efficiency.

This master thesis contains modeling, analysis and control design for atethered airfoil. A path-following controller has been developed and provenlocally asymptotically stable. The guidance law is general, and applicableto other path-following systems. The closed loop system is demonstratedin simulations, where a certain level of robustness is concluded.

The literature in the feld of multi-agent manufacturing control predom-inantly presents qualitative arguments to motivate its usage. This workpresents quantitative meaning to some of these arguments. The work alsodiscusses the control structuring process and puts it into perspective withmulti-agent control. A suitable example for the analysis of multi-agent con-trol is developed. A two layered approach for the control of this exampleprocess is proposed and it is shown that agents can be used in both of theselayers. An ad-hoc analysis is made of the top layer and simulation resultsshows that a distributed control approach introduces an optimality gap whencompared to a centralized approach. Further simulation results show thata layered approach to multi-agent control for the proposed process improvesoverall performance by increasing agent coordination.To date there has been a lack of industry adoption of multi-agent tech-nologies. The industry seems to favor control approaches that are well testedand with quantifiable evidence supporting their efficiency. The developmentof simulators that capture the full complexity of manufacturing processes isa time consuming task. The example process proposed in this work onlycaptures some core dynamics. Further work on presenting more quantifiablearguments for the use of multi-agent control is thus motivated throughoutthe thesis.The ideas presented in this thesis are included in the article ’MAS formanufacturing control: A layered case study’ by Pedersen et al. The articleis aimed for submission to the AAMAS 2012 conference in Valencia.

Gjennom denne oppgåva har det blitt laga eit system for styring av tekniske installasjonar, basert på ZigBee som kommunikasjonsteknologi. Systemet er laga slik at det òg kan nyttast som ei plattform for andre applikasjonar seinare.Arbeidet vart gjort på bakgrunn av eit studie av både eksisterande, kommersielle system og aktuelle forskingsprosjekt innanfor området. Forskingsaktiviteten er stor, og det er relativt stort spenn i prosjekta som er omtalt, frå reine tekniske prosjekt, til prosjekt med fokus meir på dei sosiale konsekvensane av meir intelligente heimar. Basert på delar av dei eksisterande systema, forskingsresultat og eigne idear, vart det laga eit bruksscenario som har fungert som ein referanse i det vidare arbeidet. Dette scenariet skildrar korleis eit slikt system kan fungera, både i seg sjølv og som ei plattform for andre applikasjonar. Som grunnleggjande konsept valde ein å laga systemet desentralisert, i det atkonfigurasjonen ligg lagra i dei enkelte einingane i systemet. Likevel har systemet ein sentral, for å kunna gjera automatisert, sentral styring. Sentralen skal òg tilby styring og konfigurasjon gjennom eit web-grensesnitt.Av fysiske einingar vart det i denne omgang laga ein sentral, ein brytarnode og ein relénode. Desse kan saman nyttast til å slå av og på alle typar elektriske einingar, opp til 16 A ved 230 V. Ein valde å skilja alle komponentar som har med ZigBee-kommunikasjon ut på eit eige kort. Alle dei tre andre nodane nyttar difor same type kort her. Å ha dette som eit eige kort, gir betre fleksibilitet, og gjer det enklare å nytta dette kortet i andre prosjekt. Tidlegare arbeid har gjort ein kjent med Atmel sine komponentar, og ein valde å nytta desse òg no, i form av Atmega128RFA1. Dette er ein mikrokontrollar med innebygd kommunikasjonsbrikke. Heimesentralen vart i tillegg til Atmel-brikka bygd rundt Beagleboard -xM, ein kraftig ARM-basert enkelt-kort-datamaskin, med nettverksfunksjonalitet, USB-portar og microSD-kort for lagring. Ein spesielt tilpassa Linux-distribusjon med tilhøyrande pakkesystem vart laga ved hjelp av Yocto-prosjektet. Dette gir eit system som startar svært raskt, og er enkelt å utvida med nye applikasjonar.Web-grensesnittet vart implementert på toppen av dette, basert på Python-rammeverket Django. Bruken av eit fullverdig operativsystem vil gjera det enkelt å integrera andre applikasjonar i denne sentralen seinare. Under utvikling og testing vart det tydeleg at rekkevidda til nodane ikkje var så god som ho burde vore. Truleg kjem dette av at impedansen i kretskortbanane mellom mikrokontrollaren og antenna ikkje er heilt riktig. Dette kan difor truleg fiksast i ein ny revisjon av ZigBee-kortet. Til gjengjeld viste batterilevetida seg å vera svært god, med håp om å få levetid på over 3 år i faktisk bruk ut av 2 AAA-batteri. Òg responstida i systemet var god, med tider på om lag 5 ms frå ein trykkar på ein knapp til reléet skifta tilstand. Å laga maskinvare tok dessverre svært mykje lenger tid enn ein hadde rekna med. Ein har difor ikkje rukke å implementera den funksjonaliteten ein hadde håpa. I tillegg finst det utfordringar i forhold til sikkerheit i nettverket, og det å få inn nye nodar i nettverket på ein trygg måte. Det er difor stort rom for vidare arbeid.Totalt sett kan systemet ikkje seiast å vera komplett, slik det framstår i dag. Primært er det noko manglande funksjonalitet, samt den dårlege rekkevidda som er dei største problema. Arbeidet som er gjort her, er likevel eit svært godt grunnlag. For det første kan nettverket og sentralen nyttast til å implementera nye applikasjonar innanfor husstyring i vid forstand. For det andre burde eit system som her òg ha eit visst kommersielt potensiale, om ein fullfører ein del av den manglande funksjonaliteten og finpussar på brukargrensesnitta.

Many unsolved problems exists in the field of robot control.This text investigates state of the art methods for path finding, trajectory generation and control in order to identify their properties, which problems they are applicable to, and their weaknesses.This is done by applying them to problems with actual real-world relevance.