Denne oppgavebesvarelsen beskriver forbedringsprosessen av en robot, samt prosessen rundt å utvide et eksisterende system for generering av kart over ukjente områder. Utvidelsen består i at systemet nå kan håndtere to ulike roboter, der det i utgangspunktet kun håndterte den ene. Den roboten som har blitt forbedret består i all hovedsak av to hjul med hver sin motor, slik at den kan bevege seg i bakkeplanet, og et sensortårn med fire infrarøde sensorer til avstandsmåling. Sensortårnet er plassert på en servomotor, slik at det kan roteres, og sensorene kan måle avstanden til omgivelsene 360° rundt roboten. For å holde orden på sin egen posisjon, er roboten utstyrt med to optiske sensorer som registrerer hjulenes omdreining, og dermed hvor langt, og i hvilken retning roboten har beveget seg. Alt dette styres av en mikrokontroller, som også holder styr på kommunikasjonen mot et brukergrensesnitt over bluetooth. Dette grensesnittet er utviklet i Matlab, og genererer et kart over omgivelsene basert på målinger fra roboten. I tillegg er det dette Matlab-systemet som forteller roboten hvor den skal bevege seg.IR-roboten led i utgangspunktet av hjulspinn, grunnet feil vektbalanse og dårlige hjul. Ved å flytte massemiddelpunktet og bytte hjul har dette problemet blitt mer eller mindre eliminert. Sammen med en ny reguleringsprosess som sammenligner de to drivhjulenes bevegelser, gjør dette at robotens bevegelsesnøyaktighet har blitt kraftig forbedret.For å bedre robotens autonome egenskaper, har den fått implementert batterimodellering. Batterimodellen simulerer robotens strømbruk, og benyttes i forbindelse med allerede eksisterende ladeautomatikk, slik at systemet selv vet når roboten trenger å lades. Denne utvidelsen av systemet har fjernet det siste ikke-autonome momentet i kartleggingsprosessen, slik at det nå i prinsippet bare er å trykke på start-knappen i brukergrensesnittet, så holder roboten på til kartleggingen er ferdig.Totalt sett har IR-roboten gjennomgått mange små forbedringer, som til sammen utgjør en vesentlig forbedring, og oppgavetekstens mål om forbedringer av IR-roboten kan sies å være nådd.Den andre roboten er konstruert på tilsvarende vis som IR-roboten. Den har to motorer med hvert sitt hjul, i tillegg til et sensortårn som kan roteres. Tårnet består i dette tilfellet av et trådløst webkamera, samt en avstandssensor basert på ultrasonisk lyd. Roboten observerer sine omgivelser ved ved å undersøke bilder fra kameraet, og sjekker avstanden til det den finner med ultralydsensoren. Oppgavebesvarelsen viser videre hvordan kameraroboten har blitt inkludert i IR-robotens grensesnitt, slik at dette grensesnitte kan kontrollere begge robotene. Systemet har blitt tilpasset slik at måledata fra kamera-roboten også fungerer i kartgenereringsprosessen. I utgangspunktet var meningen å la grensesnittet operere begge robotene samtidig, men dette er ikke gjort. Dermed vil et naturlig neste trinn være å fortsette utvidelsen av systemet til å fungere med to eller flere roboter parallelt. Arbeidet som er gjort for å utvide systemet til to roboter skal ha lagt et godt grunnlag for å kunne ta utvidelsen et nytt steg videre.

The key objective of this work is the proposition of the path planning strategy for unmanned aerial vehicle (UAV) intended for information gathering in Arctic environments / ice-infested regions. Two different path planning strategies are considered; one for analysis of the surveillance area defined as a grid and the other for analysis of the surveillance area defined as a sector. The mixed integer linear programming (MILP), YALMIP modeling language and GUROBI optimizer are used for formulation and solution of the path planning optimization problems. Furthermore, both path planning strategies are tested for the cases of constant and variable ice flow, respectively; the following are investigated in each simulation case: flight path of the UAV, coverage of the surveillance area, speed and acceleration of the UAV and the solver-time. Moreover, throughout this work only a planar motion is considered, with one single UAV used in each simulation case.

This thesis has studied Haraldrud MSW combustion process. Haraldrud is a realcombustion plant burning waste for citizens of Oslo. A thoroughly description ofthe combustion process has been presented based on manipulating and analyzing theprocess, together with long discussions with the plant engineers. Working with a realplant is time-consuming, challenging and very informative. Rarely theories are easyto implement on a real plant, and the focus of this thesis has been to connect theoryand practice. All simulations are based on real process data.Burning MSW is a complex process to control, and several factors including;unknown calorific value, regulation of the waste flow, and the long time constant formeasuring and regulating the energy contributes to this. Today’s unknown calorificvalue and measuring the energy from the combustion can be calculated from flue gasmeasurements. By implementing these in a new controller at Haraldrud the variationin energy from the combustion will be reduced and results in waste flow increase.One method to develop a model and estimate the model parameters is to decide amodel set and estimate the model parameters based on system identification theory.An open-loop system identification test was applied to the combustion process. Further,the model parameters to a MIMO ARX model were estimated from recordedtest data. It was concluded that the process variation in the gathered open-loopdata was too large, which resulted in a poor model. The model should instead beestimated based on closed-loop identification.Another method to estimate the model parameters is to use a control law toestimate the parameters on-line. Gradient method is one control law and has beenvalidated from recorded process data. The simulation shows that the estimated andmeasured outputs followed each other perfectly.

In order to efficiently control an unmanned vehicle, knowledge about the position, velocity and attitude (orientation) is needed. This thesis address this problem, and designs a navigation system for local navigation using low-cost sensors. Two loosely coupled GNSS/IMU integration filters are developed using a direct state estimation approach. The first is a quaternion based multiplicative extended Kalman filter (MEKF). A multiplicative filter differs from the usual EKF in how the attitude is represented, which is done by a quaternion product. The filter avoids the singular covariance matrix caused by the constraint on the quaternion. Two versions of the filter are developed: one using the q-method to get a measurement of the attitude; and one using vector measurements directly. The second is a nonlinear observer, termed HuaMahony. It is derived by combining two nonlinear algorithms proposed by Mahony et al. and Hua. The resulting nonlinear observer is able to estimate the linear acceleration as well as gyro bias. The nonlinear observer is written on an EKF-like discrete-time corrector-predictor formulation.Both the multiplicative extended Kalman filter and the nonlinear observer are tested and verified through simulations and experimental data. Tests are carried out to examine how different disturbances affects the estimates and to compare the performance results. Simulation results shows an average dynamic accuracy of <0.5 deg RMS for the attitude estimates, for both observers. The results from the experimental tests shows an average roll, pitch and yaw accuracy of (0.3 0.3 2.0) and (0.8 0.7 4.4) deg RMS for the MEKF and HuaMahony observer respectively, where it has been assumed that the Xsens MTi-G built-in EKF estimates are the true values. An advantage of the MEKF is quicker convergence from initial errors, while an advantage of the HuaMahony observer is less computational load. Results show that the HuaMahony is three times faster than the MEKF when it comes to execution time.

I denne rapporten presenteres et rammeverk for Hardware in the Loop Simulation (HILS) i forbindelse med utvikling av Unmanned Aerial Vehicle (UAV) styresystemer. Oppgaven er utført som masteroppgave ved Teknisk Kybernetikk, NTNU.Rammeverket er utviklet i Windows 7 og baserer seg på dynamisk simulator programmert i MATLAB/Simulink og Flight Gear er brukt for visuell fremstilling av flyet. Resultatet består av tre deler som til sammen kompletterer en fullstendig HIL simulator. Oppgaven avgrenser seg til det datatekniske omkring utviklingen av HIL. Det vil si at den matematiske bakgrunnen for flymodeller og simuleringen ikke er utledet.Kapittel 2 gir en innføring i begreper og maskinvare utviklet for Odin Recce UAV. Resultatet er deretter presentert i tre deler.I kapittel 3 er oppbyggingen av et driverbibliotek for avlesing av joystick gjennomgått i detalj for språkene: C/C++, Java, MATLAB og Simulink. En grafisk bakkestasjon for logging av data og styring av modellen utviklet i MATLAB i kapittel 4.Tilsvarende systemer er beskrevet og analysert som basis for videre utvikling som siste av resultatet tilhørende kapittel 5. Her også testprosedyrer og feilkilder redegjort for.Oppgaven presenterer en generell fremgangsmåte for HIL simulering. Rammeverket er kodet med lavest mulig kobling og høy kohesjon for at løsningen skal kunne gjenbrukes senere.Ved å tilpasse den dynamiske modellen til ønsket fysisk system kan en legge til reguleringssløyfer og kontrollsystem med mulighet for å påtrykke eventuelle feilsituasjoner – brukeren får visuell tilbakemelding på flyets oppførsel via Flight Gear og bakkestasjonen gjør det også mulig å logge sanntidsdata.Prosjektet er en del av utviklingen omkring Odin Recce D6 UAV, men denne modellen er ikke brukt spesifikt i utviklingen. Mer informasjon om Odin er på www.odin.aero.

The Unmanned Vehicle Laboratory at the Department of Engineering Cybernetics at the Norwegian University of Science and Technology has the overall goal of developing an unmanned aerial system (UAS). The unmanned aerial vehicle (UAV) to be considered is the Recce D6 delta-wing produced by Odin Aero AS. A UAS is clearly quite a complex system, and the consequences of system failure during flight are obviously potentially vital. The need for a comprehensive test setup for extensive ground testing is clearly important, and a flight simulator of the UAV is a key component of such a test setup. A mathematical model of Recce D6 is suggested from a first-principle approach and together with appurtenant actuators implemented in Matlab /Simulink. A simple motion control system with reference models and simple PID controllers are also discussed and implemented. Last, a Matlab-native graphical interface is developed to substantiate the UAV handling. The simulation model, motion control system and visualization is put together in a nearly 7 minutes long demonstration video of a fully automated flight. The video illustrates a set of aircraft maneuvers and handling issues related to take off, touch and go landings, climbing, descent, banked turn, stall and landings under gusty wind and no wind. The resulting video reveals a promising basis of a simulator environment for hardware-in-the-loop testing, but further tuning of the model parameters is required to acquire an accurate artificial representation of the aircraft.

The theory of mathematical optimization is useful within a wide range of disciplines such as science, engineering, economics and industry. Application areas have been growing steadily, driving forward the development of new effective methods. Inspired by the need for fast computational schemes in wireless sensor networks, a new optimization theory, called Fast Lipschitz, has emerged to provide effective algorithms both for distributed and centralized computations. An important property of these algorithms is that a globally optimal solution is always guranteed. In this master thesis project, a new MATLAB toolbox is developed to check wether an optimization problem is F-Lipschitz and to solve it efficiently. The difficulty is posed in verifying that a given problem is in fact F-Lipschitz. However, it is shown that under certain circumstances, this operation has a computational complexity of O(n^2) for a problem with n decision variables. The toolbox provides both a graphical interface as well as inline functions. A user guide is presented, explaining the functionalities by discussions and illustrations of example problems. Among others, a convex optimization problem of distributed detection is considered, as well as a non-convex radio power allocation problem. The novel toolbox presented in this thesis may be of considerable utility in solving optimization problems and studying their characteristics.

Spacecrafts, especially satellites, play an ever greater rolein our daily lives as we increasingly depend on the services they provide,which in turn, more often than not, critically depend on maintainingcorrect payload attitude. As smaller educational satellites pavethe way for organization, group and privately owned pico-satellites, weexplore the possibilities of attitude control through magnetic coil actuation.We approach the whole problem, from control theory developmentto first prototype actualization and control algorithm implementation,presenting the steps taken in a user-friendly manner while pointing outthe pitfalls and drawbacks of different solutions. The control is based ona dissipative detumbling controller which after the initial phase is overriddenby the reference controller attaining final desired payload attitude.We find that a simple 8-bit, 16Mhz microcontroller unit has the su-cient processing power to continuously compute the geomagnetic fieldusing the complex International Geomagnetic Reference Field model,while simultaneously maintaining correct coil actuation. The powerconsumed by the controllers during the <300 minute control phase,from initial tumbling to desired attitude, given a typical tumbling velocityof absolute magnitude 0.2[rad/sec], is found to be no more than 150Joules across the randomly selected test scenarios. Thus we are able toconclude that three perpendicular magnetic coils, together with constantlypresent disturbances and complex geomagnetic field model preventingit from remaining at an ill-aligned attitude where one actuatingdegree of freedom is lost, provide sufficient actuation for reference controlof a spacecraft.

As a part of the CUBESAT project at the Norwegian University ofScience and Technology (NTNU), this paper studies development andcomparison of attitude estimation methods for small cube satellites usinglow-cost sensors. The algorithms are based on data from two vectorizedmeasurements as well as a gyroscope.In this paper a new method for attitude estimation has been de-veloped based on QUaternion ESTimation (QUEST). A major concernwith QUEST is that it cannot handle non-vectorized measurements suchas gyroscope data. Substantial improvements have been made to fusevectorized and non-vectorized measurements, making the new ExtendedQUaternion ESTimation (EQUEST) more suitable for attitude estima-tion. The well known Extended Kalman Filter (EKF) is derived andimplemented for comparison. Both methods have been developed andsimulated in MATLAB. The code have been rewritten using C language.The methods are compared both theoretically and experimentally withimplementation and testing on an AVR microcontroller. Minimum powerusage and number of arithmetic operations were considered during thesoftware development.Testing indicates that the EKF provides a smoother estimation thanthe newly developed EQUEST. In contrast to EQUEST, the EKF isable to estimate the magnetometer and accelerometer bias. However, theEQUEST has a signicantly faster settling time and is less computationalcostly. Compared to the EKF, EQUEST runs more than 5 times faster.It also requires only 8% of the arithmetic operations of the EKF. Anotherdisadvantage with the EKF is tracking problems that occur when the twovectorized measurements are close to parallel. With vectors close to paral-lel, the mathematical formulation of the EKF makes tracking of a rotationaround the parallel axis extremely dicult. These diculties are hardlyobserved in the EQUEST algorithm, which makes it very attractive forattitude estimation.The attitude control of CUBESATs is often done by magnetorquers,which will aect the local magnetic eld. Hence, control and estimationshould not be done simultaneously, resulting in the estimation and controlswitching on and o. For this reason, the long settling time of the EKFmakes the EQUEST even more attractive.The results in this paper indicate that the newly developed EQUESTis highly suitable for projects with either limited budget, space, weight orcomputational power.

Dynamic modeling means deriving equations that explicitly describes the relationship between force and motion in a system. To be able to control a robot manipulator as required by its operation, it is important to consider the dynamic model in design of the control algorithm and simulation of motion. In general there are two approaches available; the Euler-Lagrange formulation and the Newton-Euler formulation. This thesis explains briefly the differences of the formulations, and then research the Newton-Euler method in detail. A complete derivation of the method is derived, and an automated framework for applying the method on any serial manipulator with revolute joints is presented. By using the framework, the Newton-Euler formulation is applied on a modern industrial manipulator with six degrees of freedom. The dynamic parameters of the system are estimated, and the validity of the resulting dynamic model is verified by several simulations in open and closed loop.The simulations show that the system is unstable in open loop, and that it achieves global asymptotic stability in closed loop with gravity compensation, by including PD controllers with independent joint control. The latter is a confirmation of a mathematical proof based on a Lyapunov stability analysis, which is presented as well. Equivalent simulations of the dynamic model of the same manipulator derived by the standard Euler-Lagrange formulation show that the efficiency of recursive procedures is way higher that treating the manipulator as a whole.A suggestion for future work is performing thorough dynamic parameter identification. An improved model can ultimately be implemented in the controller of the manipulator, and optimized for a specific job task.