No description available.

Collision avoidance is an essential capability for autonomous and assisted-driving ground vehicles. In this work, we developed a novel model predictive control based intelligent collision avoidance (CA) algorithm for a multi-trailer industrial ground vehicle implemented on a General Purpose Graphical Processing Unit (GPGPU). The CA problem is formulated as a multi-objective optimal control problem and solved using a limited look-ahead control scheme in real-time. Through hardware-in-the-loop-simulations and experimental results obtained in this work, we have demonstrated that the proposed algorithm, using NVIDA’s CUDA framework and the NVIDIA Jetson TX2 development platform, is capable of dynamically assisting drivers and maintaining the vehicle a safe distance from the detected obstacles on-thely. We have demonstrated that a GPGPU, paired with an appropriate algorithm, can be the key enabler in relieving the computational burden that is commonly associated with model-based control problems and thus make them suitable for real-time applications.

This thesis investigates one option for improving the control of a reheating furnace used in heating steel slabs before hot rolling; an essential part of the steel manufacturing process. The furnace consumes a significant amount of energy, leading to high cost and high carbon dioxide emissions. The proposed solution is the implementation of a model predictive control (MPC) system to improve control and reduce fuel usage. The MPC system will be based on the use of system identification techniques to find a prediction model of the furnace, specifically using ARMAX models. An additional simulation model will be used to simulate the system, and to compare the performance of MPC and PID. The prediction model is found to have a normalized root mean squared error of over 91% for the first five minutes, suggesting that it has potential to be used for MPC. The simulation model has significant inaccuracies, due to the presence of unmeasured disturbances. The simulation results, although limited due to the inaccuracies of the simulation model, suggest that MPC is a viable option for improved control of the furnace. The use of MPC can potentially improve the repeatability of the heating process, resulting in improved steel quality and reduced defects. This thesis suggests that further investigation into the use of MPC for controlling reheating furnaces in the steel industry is worth pursuing, and could potentially bring significant benefits to both producers and the environment.

京都大学 / 新制・課程博士 / 博士(情報学) / 甲第24266号 / 情博第810号 / 新制||情||136(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 大塚 敏之, 教授 石井 信, 教授 森本 淳 / 学位規則第4条第1項該当 / Doctor of Informatics / Kyoto University / DFAM

This thesis presents a novel architecture for human-in-the-loop control of multiple drones. The design of such systems must address several challenges at the same time. The drones must avoid collisions with each other and with obstacles in their task environment while following operator's command as closely as possible to navigate their environment. To this end, they should be able to adjust their pre-defined desired formation and, if needed, transition to alternative formations to ensure collision-free operation in their task environment while following the operator's commands. The proposed control strategy is a central algorithm with multiple stages and relies on formulating and solving convex optimization problems in real time to achieve the control objectives. The operator provides reference velocity commands for the flock of drones to move them in the task environment. The algorithm creates linear collision avoidance constraints and distributes the operator's commands among the drones through a number of intermediate steps. It generates reference trajectories for the drones motion by solving a model-based optimization problem over a receding horizon. Conventional trajectory controllers generate the control inputs for individual drones. Prospective formation shapes are obtained for the drones by formulating and solving parallel convex optimizations, considering the operator's reference command and the obstacle-free space. While keeping the convexity of the optimization problem, the proposed algorithm allows for the presence of obstacles in the middle of the formation. This is achieved by properly assigning obstacle-free regions to each agent separately in the formation. In addition, safe convex regions in the form of linear inequality constraints are generated in the direction of the operator's commanded velocity. Moreover, constraints are introduced to avoid inter-drone collisions at each step. Trajectory optimization is formulated as a quadratic programming problem similar to model predictive control schemes to minimize deviation from human operator's command. The effectiveness of the proposed control algorithm is initially verified by simulating two different operational scenarios. Furthermore, the algorithm is implemented on actual hardware to operate a flock of three drones in a laboratory setting. The implementation of the algorithm in C++ utilizes high-performance computation techniques to achieve sufficiently high real-time control update rates for smooth and stable operation of the drones. / Thesis / Master of Applied Science (MASc) / The rise of unmanned aerial vehicle technology and the increase in their accessibility have made them viable solutions for serious missions such as search and rescue operations. Complex cooperative tasks can be conducted via a collection of drones which can show higher levels of robustness and agility as a system. Although repetitive and simple actions can be easily automated, real-world problems are unpredictable in which complex decision-making is involved. Such scenarios can be tackled by the presence of a human supervisor to empower the system with strong cognitive capabilities. This thesis presents a multi-layer control framework for human-in-the-loop operation of a flock of unmanned aerial vehicles. This method continuously optimizes the drones trajectories to adhere as closely as possible to operator's motion commands while avoiding collisions among them and with obstacles in their task environment. This new control framework is successfully validated in both simulations and experiments in a laboratory environment.

No description available.

There is a constant demand to use more advanced control methods in a wider field of applications. Model Predictive Control (MPC) is one such control method, based on recurrently solving an optimization problem for determining the optimal control signal. To solve an optimization problem can be a complex task, and it is difficult to determine beforehand how long time it will take. For a high-speed application with limited computational power, it is necessary to have an efficient algorithm to solve the optimization problem and an accurate estimation of the longest solution time. Recent research has given methods both to solve quadratic programs efficiently and to find an upper limit on the solution times. These methods are in this thesis applied to a control system based on linear MPC for the Crazyflie 2.0 nano quadcopter. The implementation is made completely online on the processor of the quadcopter, with limited computational power. A problem with the size of 36 optimization variables and 60 constraints is solved at a frequency of 100 Hz on the quadcopter. Apart from implementing MPC, a framework for computing an upper limit to the solution time has been tested. This gives a possibility to certify the formulation for real-time applications up to a well-defined maximum frequency. An implementation is shown where the framework has been used in practice to control a quadcopter flying with a real-time certified implementation of MPC. / Det finns en ständig efterfrågan för mer avancerade metoder för reglering. Modellprediktiv reglering (MPC) är en sådan avancerad metod som kräver att ett optimeringsproblem löses varje gång en ny styrsignal ska beräknas. Att lösa optimeringsproblem kan vara en komplicerad uppgift, och det är svårt att på förhand veta hur lång beräkningstid som krävs. För att MPC ska kunna användas i tillämpningar i hög hastighet och med begränsad beräkningskraft är det nödvändigt att ha en effektiv lösningsalgoritm, och även en korrekt uppskattning av den längsta lösningstiden som behövs. Aktuell forskning har gett metoder både för att effektivt lösa kvadratiska optimeringsproblem, samt för att kunna hitta en övre gräns på beräkningstiden. I den här rapporten appliceras dessa metoder på ett styrsystem baserat på MPC i en Crazyflie 2.0, vilket är en nanodrönare. Styrsystemet är implementerat helt och hållet på drönarens processor, med den begränsade datorkraft som det innebär. Ett problem med en storlek på 36 optimeringsvariabler och 60 bivillkor lösesmed en frekvens på 100 Hz. Förutom att implementera MPC har även en metod för att bestämma en övre gräns på beräkningstiden testats. Det ger en möjlighet att certifiera styrstytemetför att garanterat kunna beräkna en ny styrsignal inom den övre tiden, vilket i sin tur innebär att styrsytemet kan certificeras för realtidsanvändning i långsammare frekvenser än den övre gränsen. I rapporten visas en certifierad implementation, och data från flygning med en certifierad regulator finns med i resultatet.

This dissertation aims to reduce urban traffic congestion with street traffic signal control. The traffic signal controllers in the U.S. follow the National Electrical Manufacturing Association Standards (NEMA Standards). In a NEMA controller, the control parameters for a coordinated control are cycle, green splits, and offset. This dissertation proposed a virtual phase-link concept and developed a macroscopic model to describe the dynamics of a traffic network. The coordinated optimal splits control problem was solved using model predictive control. The outputs of the solution are the green splits that can be used in NEMA controllers. I compared the proposed method with a state-of-the-practice signal timing software under coordinated-actuated control settings. It was found that the proposed method significantly outperformed the benchmarking method. I compared the proposed NEMA-based virtual phase-link model and a Max Pressure controller model using Vissim. It was found that the virtual phase-link method outperformed two control strategies and performed close, but not as good as, the Max Pressure control strategy. The disadvantage of the virtual phase-link method stemmed from the waste of green time during a fixed control cycle length and the delay which comes from the slowing down of platoon during a road link to allow vehicles to switch lanes. Compared to the Max Pressure control strategy, the virtual phase-link method can be implemented by any traffic controller that follows the NEMA standards. The real-time requirement of the virtual phase-link method is not as strict as the Max Pressure control strategy. I introduced the offsets optimization into the virtual phase-link method. I modeled the traffic arrival pattern based on the optimization results from the virtual phase-link control method. I then derived a phase delay function based on the traffic arrival pattern. The phase delay function is a function of the offset between two consecutive intersections. This phase delay function was then used for offsets optimization along an arterial. I tested the offsets optimization method against a base case using microscopic simulations. It was found that the proposed offset optimization method can significantly reduce vehicle delays. / Doctor of Philosophy / The goal of this work is to reduce traffic congestion by providing optimized signal timing plans to controllers. Knowing that the controllers in the U.S. follow National Electrical Manufacturing Association (NEMA) Standards, I proposed a virtual phase-link concept and modeled the road traffic network under NEMA controllers’ control as a set of virtual phase-links. Each virtual phase-link corresponds to a NEMA phase at an intersection. I then proposed a NEMA-based virtual phase-link street traffic model. The control variables are the green time allocated to each phase. I compared the proposed NEMA-based virtual phase-link control method with a state-of-the-practice signal timing software using simulation experiments. It was found that the proposed control methods significantly outperformed the signal timing software. I implemented a state-of-the-art adaptive control strategy, Max Pressure control. I compared the proposed NEMA-based virtual phase-link control method with the Max Pressure control strategy. I found that the virtual phase-link control method performed close, but not as good as, the Max Pressure control strategy. The disadvantage of the virtual phase-link method stemmed from the waste of green time during a fixed control cycle length and the delay which comes from the slowing down of platoon during a road link to allow vehicles to switch lanes. The Max Pressure control needs non-conventional controllers which can potentially switch to any phase at any time. Compared to the Max Pressure control strategy, the virtual phase-link method can be implemented by any traffic controller that follows the NEMA standards. The real-time requirement of the virtual phase-link method is not as strict as the Max Pressure control strategy. I then augmented the virtual phase-link method with optimal offsets control. The offsets are the time differences of the coordinated phases comparing to a reference point in a control cycle. I derived a phase delay function and used that function to optimize the offsets by minimizing the associated delays. The simulation experiments showed that the proposed offsets optimization method could reduce the delay along the coordinated path significantly.

Traditional generators have been decommissioned or replaced by renewable energy generation due to utility long-standing goals. However, instead of flattening the entire plant, the rotating mass of generator can be utilized as a storage unit (inertia resource) to mitigate the frequency swings during transient caused by the renewables. The goal of this work is to design a control strategy utilizing the decommissioned generator interfaced with power grid via a back-to-back converter to provide inertia support. This is referred to as decoupled synchronous machine system (DSMS). On top of that, the grid-side converter is capable of providing reactive power as an auxiliary voltage controller. However, in a practical setting, for power utilities, the detailed state equations of such device as well as the complicated nonlinear power system are usually unobtainable making the controller design a challenging problem. Therefore, a model free, purely data-driven strategy for the nonlinear controller design using Koopman operator-based framework is proposed. Besides, the time delay embedding technique is adopted together with Koopman operator theory for the nonlinear system identification. Koopman operator provides a linear representation of the system and thereby the classical linear control algorithms can be applied. In this work, model predictive control is adopted to cope with the constraints of the control signals. The effectiveness and robustness of the proposed system are demonstrated in Kundur two-area system and IEEE 39-bus system. / Doctor of Philosophy / Power system is facing an energy transformation from the traditional fuel to sustainable renewable such as solar, wind and so on. Unlike the traditional fuel energized generators, the renewable has very little inertia to maintain frequency stability. Therefore, this work proposes a new system referred to as decoupled synchronous machine system (DSMS) to support the grid frequency. DSMS consists of the rotating mass of generator and a back-to-back converter which can be utilized as an inertia resource to mitigate the frequency oscillations. In addition, the grid-side converter can provide reactive power to improve voltage performance during faults. This work aims to design a control strategy utilizing DSMS to support grid frequency and voltage. However, an explicit mathematical model of such device is unobtainable in a practical setting making data-driven control the only option. A data-driven technique which is Koopman operator-based framework together with time delay embedding algorithm is proposed to obtain a linear representation of the system. The effectiveness and robustness of the proposed system are demonstrated in Kundur two-area system and IEEE 39-bus system.

Spacecraft relative motion modeling and control promises to enable or augment a wide range of missions for scientific research, military applications, and space situational awareness. This dissertation focuses on the development of novel, optimization-based, control design for some representative relative-motion-enabled missions. Spacecraft relative motion refers to two (or more) satellites in nearly identical orbits. We examine control design for relative configurations on the scale of meters (for the purposes of proximity operations) as well as on the scale of tens of kilometers (representative of science gathering missions). Realistic control design for satellites is limited by accurate modeling of the relative orbital perturbations as well as the highly constrained nature of most space systems. We present solutions to several types of optimal orbital maneuvers using a variety of different, realistic assumptions based on the maneuver objectives. Initially, we assume a perfectly circular orbit with a perfectly spherical Earth and analytically solve the under-actuated, minimum-energy, optimal transfer using techniques from optimal control and linear operator theory. The resulting open-loop control law is guaranteed to be a global optimum. Then, recognizing that very few, if any, orbits are truly circular, the optimal transfer problem is generalized to the elliptical linear and nonlinear systems which describe the relative motion. Solution of the minimum energy transfer for both the linear and nonlinear systems reveals that the resulting trajectories are nearly identical, implying that the nonlinearity has little effect on the relative motion. A continuous-time, nonlinear, sliding mode controller which tracks the linear trajectory in the presence of a higher fidelity orbit model shows that the closed-loop system is both asymptotically stable and robust to disturbances and un-modeled dynamics. Next, a novel method of computing discrete-time, multi-revolution, finite-thrust, fuel-optimal, relative orbit transfers near an elliptical, perturbed orbit is presented. The optimal control problem is based on the classical, continuous-time, fuel-optimization problem from calculus of variations, and we present the discrete-time analogue of this problem using a transcription-based method. The resulting linear program guarantees a global optimum in terms of fuel consumption, and we validate the results using classical impulsive orbit transfer theory. The new method is shown to converge to classical impulsive orbit transfer theory in the limit that the duration of the zero-order hold discretization approaches zero and the time horizon extends to infinity. Then the fuel/time optimal control problem is solved using a hybrid approach which uses a linear program to solve the fuel optimization, and a genetic algorithm to find the minimizing time-of-flight. The method developed in this work allows mission planners to determine the feasibility for realistic spacecraft and motion models. Proximity operations for robotic inspection have the potential to aid manned and unmanned systems in space situational awareness and contingency planning in the event of emergency. A potential limiting factor is the large number of constraints imposed on the inspector vehicle due to collision avoidance constraints and limited power and computational resources. We examine this problem and present a solution to the coupled orbit and attitude control problem using model predictive control. This control technique allows state and control constraints to be encoded as a mathematical program which is solved on-line. We present a new thruster constraint which models the minimum-impulse bit as a semi-continuous variable, resulting in a mixed-integer program. The new model, while computationally more expensive, is shown to be more fuel-efficient than a sub-optimal approximation. The result is a fuel efficient, trajectory tracking, model predictive controller with a linear-quadratic attitude regulator which tracks along a pre-computed ``safe'' trajectory in the presence of un-modeled dynamics on a higher fidelity orbital and attitude model. / Ph. D.