This compilation thesis presents an overarching framework on the utilization of nonlinear model predictive control(NMPC) for various applications in the context of Unmanned Aerial Vehicle (UAV) path planning and collision avoidance. Fast and novel optimization algorithms allow for NMPC formulations with high runtime requirement, as those posed by controlling UAVs, to also have sufficiently large prediction horizons as to in an efficient manner integrate collision avoidance in the form of set-exclusion constraints that constrain the available position-space of the robot. This allows for an elegant merging of set-point reference tracking with the collision avoidance problem, all integrated in the control layer of the UAV. The works included in this thesis presents the UAV modeling, cost functions, constraint definitions, as well as the utilized optimization framework. Additional contributions include the use case on multi-agent systems, how to classify and predict trajectories of moving (dynamic) obstacles, as well as obstacle prioritization when an aerial agent is in the precense of more obstacles, or other aerial agents, than can reasonably be defined in the NMPC formulation. For the cases of dynamic obstacles and for multi-agent distributed collision avoidance this thesis offers extensive experimental validation of the overall NMPC framework. These works push the limits of the State-of-the-Art regarding real-time real-life implementations of NMPC-based collision avoidance. The works also include a novel RRT-based exploration framework that combines path planning with exploration behavior. Here, a multi-path RRT * planner plans paths to multiple pseudo-random goals based on a sensor model and evaluates them based on the potential information gain, distance travelled, and the optimimal actuation along the paths.The actuation is solved for as as the solutions to a NMPC problem, implying that the nonlinear actuator-based and dynamically constrained UAV model is considered as part of the combined exploration plus path planning problem. To the authors best knowledge, this is the first time the optimal actuation has been considered in such a planning problem. For all of these applications, the utilized optimization framework is the Optimization Engine: a code-generation framework that generates a custom Rust-based solver from a specified model, cost function, and constraints. The Optimization Engine solves general nonlinear and nonconvex optimization problems, and in this thesis we offer extensive experimental validation of the utilized Proximal-Averaged Newton-type method for Optimal Control (PANOC) algorithm as well as both the integrated Penalty Method and Augmented Lagrangian Method for handling the nonlinear nonconvex constraints that result from collision avoidance problems.

Today, the control of heat pumps aims to first and foremost maintain a comfortable indoor temperature. This is primarily done by deciding input power based on outside temperature. The cost of electricity, which can be rather volatile, is not taken into account. Electricity price can be provided on an hourly rate, and since a house can store thermal energy for a duration of time, it is possible to move electricity consumption to hours when electricity is cheap. In this thesis, the strategy used in the developed controller is Model Predictive Control (MPC). It is a suitable strategy because of the ability to incorporate an objective function that can be designed to take the trade-off between indoor temperature and electricity cost into account. The MPC prediction horizon is dynamic as the horizon of known electricity spot prices varies between 12 and 36 hours throughout the day. We model a residential house heated with a ground source heat pump for use in a case analysis. Sampled weather and spot price data for three different weeks are used in computer simulations. The developed MPC controller is compared with a classic \textit{heat curve} controller, as well as with variations of the MPC controller to estimate the effects of prediction and model errors.  The MPC controller is found to be able to reduce the electricity cost and/or provide better comfort and the prioritization of these factors can be changed depending on user preferences. When shifting energy consumption in time it is necessary to store thermal energy somewhere. If the house itself is used for this purpose, variations in indoor temperature must be accepted. Further, accurate modeling of the Coefficient of Performance (COP) is essential for ground source heat pumps. The COP varies significantly depending on operating conditions and the MPC controller must therefore have a correct perception of the COP. Publicly available weather forecasts are of sufficient quality to be usable for future prediction of outside temperature. For future studies, it would be advantageous if better models can be developed for prediction of global radiation. Including radiation in the MPC controller model would enable better comfort with very similar operating costs compared to when the MPC controller does not take radiation into account.

Truck platooning is a technology where trucks drive in a formation with each other with a small distance in between trucks in order to save fuel and reduce emissions. In this project,a distributed method for solving the optimal time problem for every truck in a hub-based transport system will be developed.Each truck will have its own utility function to optimize and is able to adjust its schedule independently. To create and test the method, a simulation of hundreds of trucks in a network of routes was created using the Python language. The results produced by running the simulation were positive and realistic. / Konvojkörning med lastbilar är en teknologi där lastbilar kör i en formation med varandra med små avstånd mellan lastbil för att spara på bränsle och minska utsläppen. I det här projektet kommer en distribuerande metod för att lösa det optimala tidsschemat för varje lastbil i ett navbaserat tranportsystem att utvecklas. Varje lastbil kommer att ha sin egen vinstfunktion att optimera och kommer självständigt att kunna ändra sitt reseschema. För att skapa och testa metoden kördes en simulation som skrevs i Python, och som behandlade hundratals lastbilar i ett nätverk av vägar. Resultaten som simulationen cerade var positiva och realistiska. / Kandidatexjobb i elektroteknik 2022, KTH, Stockholm

In this dissertation, we focus on the strategy that places and stabilizes the path of an aerial drogue, which is towed by a mothership aircraft using a long flexible cable, onto a horizontally flat orbit by maneuvering the mothership in the presence of wind. To achieve this goal, several studies for towed cable systems are conducted, which include the dynamic modeling for the cable, trajectory generation strategies for the mothership, trajectory-tracking control law design, and simulation and flight test implementations. First, a discretized approximation method based on finite element and lumped mass is employed to establish the mathematical model for the towed cable system in the simulation. Two approaches, Gauss's Principle and Newton's second law, are utilized to derive the equations of motion for inelastic and elastic cables, respectively. The preliminary studies for several key parameters of the system are conducted to learn their sensitivities to the system motion in the steady state. Flight test results are used to validate the mathematical model as well as to determine an appropriate number of cable links. Furthermore, differential flatness and model predictive control based methods are used to produce a mothership trajectory that leads the drogue onto a desired orbit. Different desired drogue orbits are utilized to generate required mothership trajectories in different wind conditions. The trajectory generation for a transitional flight in which the system flies from a straight and level flight into a circular orbit is also presented. The numerical results are presented to illustrate the required mothership orbits and its maneuverability in different wind conditions. A waypoint following based strategy for mothership to track its desired trajectory in flight test is developed. The flight test results are also presented to illustrate the effectiveness of the trajectory generation methods. In addition, a nonlinear time-varying feedback control law is developed to regulate the mothership to follow the desired trajectory in the presence of wind. Cable tensions and wind disturbance are taken into account in the design model and Lyapunov based backstepping technique is employed to develop the controller. The mothership tracking error is proved to be capable of exponentially converging to an ultimate bound, which is a function of the upper limit of the unknown component of the wind. The simulation results are presented to validate the controller. Finally, a trajectory-tracking strategy for unmanned aerial vehicles is developed where the autopilot is involved in the feedback controller design. The trajectory-tracking controller is derived based on a generalized design model using Lyapunov based backstepping. The augmentations of the design model and trajectory-tracking controller are conducted to involve the autopilot in the closed-loop system. Lyapunov stability theory is used to guarantee the augmented controller is capable of driving the vehicle to exponentially converge to and follow the desired trajectory with the other states remaining bounded. Numerical and Software-In-the-Loop simulation results are presented to validate the augmented controller. This method presents a framework of implementing the developed trajectory-tracking controllers for unmanned aerial vehicles without any modification to the autopilot.

In this work, an improved Learning Model Predictive Control (LMPC)architecture for autonomous racing is presented. The controller is referencefree and is able to improve lap time by learning from history data of previouslaps. A terminal cost and a sampled safe set are learned from history data toguarantee recursive feasibility and non-decreasing performance at each lap.Improvements have been proposed to implement LMPC on autonomousracing in a more efficient and reliable way. Improvements have been doneon three aspects. Firstly, system identification has been improved to be runin a more efficient way by collecting feature data in subspace, so that thesize of feature data set is reduced and time needed to run sorting algorithmcan be reduced. Secondly, different strategies have been proposed toimprove model accuracy, such as least mean square with/without lifting andGaussian process regression. Thirdly, for reducing algorithm complexity,methods combining different model construction strategies were proposed.Also, running controller in a multi-rate way has also been proposed toreduced algorithm complexity when increment of controller frequency isnecessary. Besides, the performance of different system identificationstrategies have been compared, which include strategy from newton’s law,strategy from classical system identification and strategy from machinelearning. Factors that can possibly influence converged result of LMPCwere also investigated, such as prediction horizon, controller frequency.Experiment results on a 1:10 scaled RC car illustrates the effectiveness ofproposed improvements and the difference of different system identificationstrategies. / I detta arbete, presenteras en förbättrad inlärning baserad modell prediktivkontroll (LMPC) för autonom racing, styralgoritm är referens fritt och har visatsig att kunna förbättra varvtid genom att lära sig ifrån historiska data från tidigarevarv. En terminal kostnad och en samplad säker mängd är lärde ifrån historiskdata för att garantera rekursiv genomförbarhet och icke-avtagande prestanda vidvarje varv.förbättringar har presenterats för implementering av LMPC på autonom racingpå ett mer effektivt och pålitligt sätt. Förbättringar har gjorts på tre aspekter.Först, för system identifiering, föreslår vi att samlar feature data i delrummet,så att storlek på samlade datamängd reduceras och tiden som krävs för attköra sorteringsalgoritm minskas. För det andra, föreslår vi olika strategierför förbättrade modellnoggrannheten, såsom LMS med/utan lyft och Gaussianprocess regression. För det tredje, För att reducerar komplexitet för algoritm,metoder som kombinerar olika modellbygg strategier föreslogs. Att körastyrenhet på ett multi-rate sätt har också föreslagits till för att reduceraalgoritmkomplexitet då inkrementet av styrfrekvensen är nödvändigt.Prestanda av olika systemidentifiering har jämförts, bland annat, Newtonslag, klassisk systemidentifierings metoder och strategier från maskininlärning.Faktorer som eventuellt kan påverka konvergens av LMPC resultat har ocksåundersökts. Såsom, prediktions horisont, styrfrekvensen.Experimentresultat på en 1:10 skalad RC-bilen visar effektiviteten hos föreslagnaförbättringarna och skillnaderna i olika systemidentifierings strategier.

京都大学 / 新制・課程博士 / 博士(情報学) / 甲第24040号 / 情博第796号 / 新制||情||135(附属図書館) / 京都大学大学院情報学研究科システム科学専攻 / (主査)教授 加納 学, 教授 大塚 敏之, 教授 下平 英寿 / 学位規則第4条第1項該当 / Doctor of Informatics / Kyoto University / DGAM

No description available.

The increasing market of small and affordable space systems requires fast and reliablelaunch capabilities to cover the current and future demand. This project aims to studyand implement guidance and control schemes for vertical ascent and descent phases ofa reusable launcher. Specifically, the thesis focuses on developing and applying ModelPredictive Control (MPC) and optimisation techniques to several kino-dynamic modelsof rockets. Moreover, the classical MPC method has been modified to include a decreasingfactor for the horizon used, enhancing the performance of the guidance and control.Multiple scenarios of vertical launch, landing and full fligth guidance on Earth and Mars,along with Monte Carlo analysis, were carried out to demonstrate the robustness of thealgorithm against different initial conditions. The results were promising and invite tofurther research.

Tires, being the only component of the vehicle in contact with the road surface, are responsible for generating the forces for maintaining the vehicle pose, orientation and stability of the vehicle. Additionally, the on-board advanced chassis control systems require estimation of these tire-road interaction properties for their operation. Extraction of these properties becomes extremely important in handling limit maneuvers such as Double Lane Change (DLC) and cornering wherein the lateral force transfer is dependent upon these computations. This research focuses on the development of a high-fidelity vehicle-tire model and control algorithm framework for vehicle trajectory tracking for vehicles operating in this limit handling regime. This combined vehicle-tire model places an emphasis on the lateral dynamics of the vehicle by integrating the effects of relaxation length on the contact patch force generation. The vertical dynamics of the vehicle have also been analyzed, and a novel double damper has been mathematically modeled and experimentally validated. Different control algorithms, both classical and machine learning-based, have been developed for optimizing this vertical dynamics model. Experimental data has been collected by instrumenting a vehicle with in-tire accelerometers, IMU, GPS, and encoders for slalom and lane change maneuvers. Different state estimation techniques have been developed to predict the vehicle side slip angle, tire slip angle, and normal load to further assist the developed vehicle-tire model. To make the entire framework more robust, Machine Learning algorithms have been developed to classify between different levels of tire wear. The effect of tire tread wear on the pneumatic trail of the tire has been further evaluated, which affects the aligning moment and lateral force generation. Finally, a Model Predictive Control (MPC) framework has been developed to compare the performance between the conventional vehicle models and the developed vehicle models in tracking a reference trajectory. / Doctor of Philosophy / In our rapidly advancing world, self-driving or autonomous vehicles are no longer a vision of the future but a reality of today. As we grow more reliant on these vehicles, ensuring their safety and reliability becomes increasingly critical. Unlike traditional vehicles, self-driving cars operate without human intervention. Consequently, the onus of passenger and pedestrian safety falls squarely on the vehicle's control systems. The efficiency and effectiveness of these control systems are pivotal in preventing accidents and ensuring a smooth ride. One vital aspect of these control systems lies in understanding the tires' behavior, the only parts of the vehicle that are in contact with the road surface. A tire's interaction with the road surface significantly impacts the vehicle's handling and stability. Information such as how much of the tire is in contact with the road, the forces and moments generated at this contact point, becomes valuable for optimizing the vehicle's performance. This is particularly crucial when a vehicle is turning or cornering, where the forces developed between the tires and the road are key to maintaining control and stability. In this research, a framework has been designed to improve the vehicle performance, primarily by improving the modeling of tire lag dynamics. This refers to the delay or 'lag' between a change in tire conditions (such as pressure, wear, and temperature) and the corresponding change in tire behavior. In addition, in this research a vertical dynamics model of the vehicle has also been developed incorporated with a novel double damper suspension system. To complete the entire framework, the effect of tire wear over time and how this affects its performance and safety characteristics has also been examined. By estimating and understanding this wear, we can predict how it will affect the dynamic properties of the tire, thus improving the reliability and efficiency of our autonomous vehicles. The last piece of this framework comprises the development of an MPC controller to track a reference trajectory and evaluate the performance of the developed model.

Inom reglerteknik har integrationen av maskininlärningsmetoder framträtt som en central strategi för att förbättra prestanda och adaptivitet hos styrsystem. Betydande framsteg har gjorts inom flera viktiga aspekter av reglerkretsen, såsom inlärningsbaserade metoder för systemidentifiering och parameterskattning, filtrering och brusreducering samt reglersyntes. Denna avhandling fördjupar sig i området inlärning för reglerteknik med särskild betoning på inlärningsbaserade regulatorer och identifieringsmetoder.  Avhandlingens första del behandlar undersökningen av neuronnätsbaserad Modellprediktiv Reglering (MPC). Olika nätstrukturer studeras, både generella black box-nät och nät som väver in MPC-specifik information i sin struktur. Dessa nät jämförs och utvärderas med avseende på två prestandamått genom experiment på realistiska två- och fyrdimensionella system. Den huvudsakliga nyskapande aspekten är inkluderingen av gradientdata i träningsprocessen, vilket visar sig förbättra noggrannheten av de genererade styrsignalerna. Vidare påvisar de experimentella resultaten att en MPC-informerad nätstruktur leder till förbättrad prestanda när mängden träningsdata är begränsad.  Med insikt om vikten av noggranna matematiska modeller av styrsystemet, riktar den andra delen av avhandlingen sitt fokus mot inlärningsbaserade identifieringsmetoder. Denna forskningsgren behandlar karakterisering och modellering av dynamiska system med hjälp av maskininlärning. Avhandlingen bidrar till området genom att introducera kooperativa systemidentifieringsmetoder för att förbättra parameterskattningen. Specifikt utnyttjas verktyg från Optimal Transport för att introducera en ny och mer generell formulering av ramverket Correctional Learning. Detta ramverk är baserat på en mästare-lärlingsmodell, där en expertagent (mästare) observerar och modifierar den insamlade data som används av en lärande agent (lärling), med syftet att förbättra lärlingens skattningsprocess. Genom att formulera correctional learning som ett optimal transport-problem erhålls ett mer flexibelt ramverk, bättre lämpat för skattning av komplexa systemegenskaper samt anpassning till alternativa handlingsstrategier. / In the context of control systems, the integration of machine learning mechanisms has emerged as a key approach for improving performance and adaptability. Notable progress has been made across several aspects of the control loop, including learning-based techniques for system identification and estimation, filtering and denoising, and controller design. This thesis delves into the rapidly expanding domain of learning in control, with a particular focus placed on learning-based controllers and learning-based identification methods. The first part of this thesis is devoted to the investigation of Neural Network approximations of Model Predictive Control (MPC). Model-agnostic neural network structures are compared to networks employing MPC-specific information, and evaluated in terms of two performance metrics. The main novel aspect lies in the incorporation of gradient data in the training process, which is shown to enhance the accuracy of the network generated control inputs. Furthermore, experimental results reveal that MPC-informed networks outperform the agnostic counterparts in scenarios when training data is limited. In acknowledgement of the crucial role accurate system models play in in the control loop, the second part of this thesis lends its focus to learning-based identification methods. This line of work addresses the important task of characterizing and modeling dynamical systems, by introducing cooperative system identification techniques to enhance estimation performance. Specifically, it presents a novel and generalized formulation of the Correctional Learning framework, leveraging tools from Optimal Transport. The correctional learning framework centers around a teacher-student model, where an expert agent (teacher) modifies the sampled data used by the learner agent (student), to improve the student's estimation process. By formulating correctional learning as an optimal transport problem, a more adaptable framework is achieved, better suited for estimating complex system characteristics and accommodating alternative intervention strategies. / VR 2018-03438 projekt 3224