This research focuses primary on recognition of lane-change behaviors using support vector machines (SVMs). Previous research and statistical results show that the vast majority of motor vehicle accidents are caused by driver behavior and errors. Therefore, the interpretation and evaluation of driver behavior is important for road safety analysis and improvement. The main limit to understanding driver behavior is the data availability. In particular, a full-scale lane-change data set is difficult to collect in a real traffic environment because of the safety and cost issues. Considering the data demands of the recognition model development and the obstacles of field data collection, data were collected from two aspects: simulation data and the field data. To obtain field data, an in-vehicle data recorder (IVDR) that integrates a Global Positioning System (GPS) and Inertial Measurement Unit (IMU) are developed to collect data on speed, position, attitude, acceleration, etc. To obtain simulation data, a lane-change simulation with a speed controller and a trajectory tracking controller with preview ability were developed, and sufficient lane-change data were generated. Proportional-Integral-Derivative (PID) control is applied to the speed controller and trajectory tracking controller. Simulation data were divided into two classes: dual lane-change data and single lane-change data; field data were further divided as single lane-change and non-lane-change data. Two-class and three-class classification SVM model are trained by simulation data and field data, and the model parameters were optimized by Genetic Algorithm (GA). A radial basis function and polynomial kernel functions were found that suitable for this recognition task. The recognition results indicate that, the SVM model trained by simulation data and non-lane-change data can correctly classify up to 85 percent of single lane-change field data.

Reversing with a trailer is very difficult and many drivers hesitate to even try it. This thesis examines if active steering, particularly AFS (Active Front Steering), can be used to provide assistance. For analysis and controller design a simple geometric model of car and trailer is used. The model seems to be accurate enough at the low speeds relevant for trailer reversing. It is shown that the only trailer dependent model parameter can be estimated while driving. This enables use with different trailers. Different schemes to control the system are tested. The main approach is to use the steering wheel as reference for some appropriate output signal, for example the angle between car and trailer. This makes reversing with a trailer more like reversing without a trailer. To turn left, the driver simply turns the steering wheel left and drives. Test driving, as well as theoretical analysis, shows that the resulting system is stable. Of the eight drivers that have tested this type of control, five found it to be a great advantage while two considered it more confusing than helpful. A major problem with this control approach has to do with the way AFS is constructed. With AFS, the torque required to turn the front wheels results in a reaction torque in the steering wheel. Together with the reference tracking controllers, this makes the steering wheel unstable. Theoretical analysis implies that this problem has to be solved mechanically. One solution would be to combine AFS with electric power steering. This thesis also presents a trajectory tracking scheme to autonomously reverse with a trailer. Starting from the current trailer position and the desired trajectory an appropriate turning radius for the trailer is decided. Within certain limits, this will stabilize the car as well. The desired trajectory can be programmed beforehand, but it can also be saved while driving forward. Both variants have been tested with good results.

The development of high-energy-density batteries, advanced sensor technologies, and advanced control algorithms for multirotor electric vertical takeoff and landing (eVTOL) unmanned aerial vehicles (UAVs) has led to interest in using these vehicles for a variety of applications including surveillance, package delivery, and even human transportation. In each of these cases, the ideal vehicle is one that can maneuver in congested spaces, but is efficient for traveling long distances. The combination of wings and vectored thrust make winged eVTOLs the obvious choice. However, these aircraft experience a much wider range of flight conditions that makes them challenging to model and control. This thesis contributes an aerodynamic model and a planning and control method for small, 1-2 m wingspan, winged eVTOLs. We develop the aerodynamic model based on first-principles, lumped-element aerodynamics, extending the lift and drag models to consider high-angle-of-attack flight conditions using models proposed in the literature. We present two methods for generating spline trajectories, one that uses the singular value decomposition to find a minimum-derivative polynomial spline, and one that uses B-splines to produce trajectories in the convex hull of a set of waypoints. We compare the quality of trajectories produced by both methods. Current control methods for winged eVTOL UAVs consider the vehicle primarily as a fixed-wing aircraft with the addition of vertical thrust used only during takeoff and landing. These methods provide good long-range flight handling but fail to consider the full dynamics of the vehicle for tracking complex trajectories. We present a trajectory tracking controller for the full dynamics of a winged eVTOL UAV in hover, fixed-wing, and partially transitioned flight scenarios. We show that in low- to moderate-speed flight, trajectory tracking can be achieved using a variety of pitch angles. In these conditions, the pitch of the vehicle is a free variable that we use to minimize the necessary thrust, and therefore energy consumption, of the vehicle. We use a geometric attitude controller and an airspeed-dependent control allocation scheme to operate the vehicle at a wide range of airspeeds, flight path angles, and angles of attack. We provide theoretical guarantees for the stability of the proposed control scheme assuming a standard aerodynamic model, and we present simulation results showing an average tracking error of 20 cm, an average computation rate of 800 Hz, and an 85% reduction in tracking error versus using a multirotor controller for low-speed flight.

In this thesis we study the methods for motion planning for nonholonomic systems. These systems are characterized by nonholonomic constraints on their generalized velocities. The motion planning problem with constraints on the velocities is transformed into a control problem having fewer control inputs than the degrees of freedom. The main focus of the thesis is on the study of motion planning and design of the feedback control laws for an autonomous underwater vehicle: a nonholonomic system. The nonlinear controllability issues for the system are also studied. For the design of feedback controllers, the system is transformed into chained and power forms. The methods of transforming a nonholonomic system into these forms are discussed. The work presented in this thesis is a step towards the initial study concerning the applicability of kinematic-based control on underwater vehicles. / Master of Science

In this thesis, a cognitive neural predictive controller system has been designed to guide a nonholonomic wheeled mobile robot during continuous and non-continuous trajectory tracking and to navigate through static obstacles with collision-free and minimum tracking error. The structure of the controller consists of two layers; the first layer is a neural network system that controls the mobile robot actuators in order to track a desired path. The second layer of the controller is cognitive layer that collects information from the environment and plans the optimal path. In addition to this, it detects if there is any obstacle in the path so it can be avoided by re-planning the trajectory using particle swarm optimisation (PSO) technique. Two neural networks models are used: the first model is modified Elman recurrent neural network model that describes the kinematic and dynamic model of the mobile robot and it is trained off-line and on-line stages to guarantee that the outputs of the model will accurately represent the actual outputs of the mobile robot system. The trained neural model acts as the position and orientation identifier. The second model is feedforward multi-layer perceptron neural network that describes a feedforward neural controller and it is trained off-line and its weights are adapted on-line to find the reference torques, which controls the steady-state outputs of the mobile robot system. The feedback neural controller is based on the posture neural identifier and quadratic performance index predictive optimisation algorithm for N step-ahead prediction in order to find the optimal torque action in the transient to stabilise the tracking error of the mobile robot system when the trajectory of the robot is drifted from the desired path during transient state. Three controller methodologies were developed: the first is the feedback neural controller; the second is the nonlinear PID neural feedback controller and the third is nonlinear inverse dynamic neural feedback controller, based on the back-stepping method and Lyapunov criterion. The main advantages of the presented approaches are to plan an optimal path for itself avoiding obstructions by using intelligent (PSO) technique as well as the analytically derived control law, which has significantly high computational accuracy with predictive optimisation technique to obtain the optimal torques control action and lead to minimum tracking error of the mobile robot for different types of trajectories. The proposed control algorithm has been applied to monitor a nonholonomic wheeled mobile robot, has demonstrated the capability of tracking different trajectories with continuous gradients (lemniscates and circular) or non-continuous gradients (square) with bounded external disturbances and static obstacles. Simulations results and experimental work showed the effectiveness of the proposed cognitive neural predictive control algorithm; this is demonstrated by the minimised tracking error to less than (1 cm) and obtained smoothness of the torque control signal less than maximum torque (0.236 N.m), especially when external disturbances are applied and navigating through static obstacles. Results show that the five steps-ahead prediction algorithm has better performance compared to one step-ahead for all the control methodologies because of a more complex control structure and taking into account future values of the desired one, not only the current value, as with one step-ahead method. The mean-square error method is used for each component of the state error vector to compare between each of the performance control methodologies in order to give better control results.

O planejamento de movimento é um dos problemas fundamentais em navegação autônoma para robôs móveis. Uma vez planejado o caminho, o robô executa o acompanhamento da trajetória, frequentemente, com o auxílio de um controlador em malha fechada. Este controlador tem o objetivo de minimizar os erros de acompanhamento, a fim de que a trajetória executada se aproxime da trajetória planejada. Entretanto a maioria dos planejadores de movimento não levam em consideração o modelo dinâmico do robô, dificultando assim o trabalho do controlador que executa o acompanhamento da trajetória. Incluindo as restrições cinemáticas e dinâmicas do modelo do robô, o custo computacional durante a fase de planejamento de trajetória será mais alto. Isto ocorre pois são necessárias mais variáveis para representar o espaço de estados do robô. No entanto ao levar em consideração tais restrições durante a fase de planejamento, as trajetórias geradas serão factíveis de serem acompanhadas. Os planejadores probabilísticos de movimento podem ser usados para minimizar o impacto do alto custo computacional, devido ao aumento de variáveis que representam o espaço de estados. Tais planejadores também são chamados de planejadores de movimento baseados em amostragem. A busca por um caminho livre de colisões entre dois estados é feito de maneira aleatória. Caso exista uma solução, a probabilidade do algoritmo encontrá-la tende para 1 quanto do tempo de busca tende a infinito, isto é, quanto mais tempo o algoritmo possui para realizar a busca será mais provável que ele encontre a solução. Neste trabalho é proposto um planejador de movimentos baseado em amostragem que leva em consideração os aspectos cinemáticos e dinâmicos do robô. Além disto esta abordagem de planejamento desenvolvida permite conhecer e levar em consideração os efeitos do controlador que faz o acompanhamento da trajetória, ainda na fase de planejamento de movimento. As trajetórias planejadas foram executadas no robô Pioneer 3AT. Foram levantados os dados relacionados ao desempenho do algoritmo em termos de custo computacional. E na sequência são apresentados os resultados experimentais tanto na parte de planejamento de trajetórias quanto na fase de acompanhamento. / Motion planning is one of the fundamental problems in autonomous navigation for mobile robots. Once the path is planned, the robot performs the trajectory tracking, often with the aid of a closed loop controller. This controller is designed to minimize tracking errors, in order that tracked trajectory get closer to planned path. However, the most motion planners do not take into account the dynamic model of the robot, thus hindering the work of closed loop controller. When including the kinematic constraints and dynamic model of the robot, the computational cost during the planning phase trajectory will be increased. This is because more variables are needed to represent the state space of the robot. But when taking into account these constraints during the planning phase, the trajectories generated are feasible to be followed. The probabilistic motion planners can be used to minimize the impact of high computational cost due to the increase of variables that represent the state space. These planners are also called sampling based motion planners. The search for a collision-free path between two states is done randomly. If a solution exists, the probability of the algorithm to find it tends to one while the search time tends to infinity, that is, the longer time the algorithm has to perform the search will be more likely to find the solution. This paper proposes a sampling based motion planner that takes into account the kinematic and dynamic aspects of the robot. Furthermore this approach allows one to know and take into account the effects of the controller that perform the trajectory tracking, still in the motion planning phase. The planned trajectories were performed on the robot Pioneer 3AT. Data related to the computational cost of the algorithm were analyzed. Following the experimental results are presented both in the planning of trajectories and in the tracking phase.

O planejamento de movimento é um dos problemas fundamentais em navegação autônoma para robôs móveis. Uma vez planejado o caminho, o robô executa o acompanhamento da trajetória, frequentemente, com o auxílio de um controlador em malha fechada. Este controlador tem o objetivo de minimizar os erros de acompanhamento, a fim de que a trajetória executada se aproxime da trajetória planejada. Entretanto a maioria dos planejadores de movimento não levam em consideração o modelo dinâmico do robô, dificultando assim o trabalho do controlador que executa o acompanhamento da trajetória. Incluindo as restrições cinemáticas e dinâmicas do modelo do robô, o custo computacional durante a fase de planejamento de trajetória será mais alto. Isto ocorre pois são necessárias mais variáveis para representar o espaço de estados do robô. No entanto ao levar em consideração tais restrições durante a fase de planejamento, as trajetórias geradas serão factíveis de serem acompanhadas. Os planejadores probabilísticos de movimento podem ser usados para minimizar o impacto do alto custo computacional, devido ao aumento de variáveis que representam o espaço de estados. Tais planejadores também são chamados de planejadores de movimento baseados em amostragem. A busca por um caminho livre de colisões entre dois estados é feito de maneira aleatória. Caso exista uma solução, a probabilidade do algoritmo encontrá-la tende para 1 quanto do tempo de busca tende a infinito, isto é, quanto mais tempo o algoritmo possui para realizar a busca será mais provável que ele encontre a solução. Neste trabalho é proposto um planejador de movimentos baseado em amostragem que leva em consideração os aspectos cinemáticos e dinâmicos do robô. Além disto esta abordagem de planejamento desenvolvida permite conhecer e levar em consideração os efeitos do controlador que faz o acompanhamento da trajetória, ainda na fase de planejamento de movimento. As trajetórias planejadas foram executadas no robô Pioneer 3AT. Foram levantados os dados relacionados ao desempenho do algoritmo em termos de custo computacional. E na sequência são apresentados os resultados experimentais tanto na parte de planejamento de trajetórias quanto na fase de acompanhamento. / Motion planning is one of the fundamental problems in autonomous navigation for mobile robots. Once the path is planned, the robot performs the trajectory tracking, often with the aid of a closed loop controller. This controller is designed to minimize tracking errors, in order that tracked trajectory get closer to planned path. However, the most motion planners do not take into account the dynamic model of the robot, thus hindering the work of closed loop controller. When including the kinematic constraints and dynamic model of the robot, the computational cost during the planning phase trajectory will be increased. This is because more variables are needed to represent the state space of the robot. But when taking into account these constraints during the planning phase, the trajectories generated are feasible to be followed. The probabilistic motion planners can be used to minimize the impact of high computational cost due to the increase of variables that represent the state space. These planners are also called sampling based motion planners. The search for a collision-free path between two states is done randomly. If a solution exists, the probability of the algorithm to find it tends to one while the search time tends to infinity, that is, the longer time the algorithm has to perform the search will be more likely to find the solution. This paper proposes a sampling based motion planner that takes into account the kinematic and dynamic aspects of the robot. Furthermore this approach allows one to know and take into account the effects of the controller that perform the trajectory tracking, still in the motion planning phase. The planned trajectories were performed on the robot Pioneer 3AT. Data related to the computational cost of the algorithm were analyzed. Following the experimental results are presented both in the planning of trajectories and in the tracking phase.

<p>I detta examensarbete har ett system för navigering och styrning av ett autonomt fordon implementerats. Syftet med detta arbete är att vidareutveckla fordonet som ska användas vid utvärdering av banplaneringsalgoritmer och studier av andra autonomifunktioner. Med hjälp av olika sensormodeller och sensorkonfigurationer går det även att utvärdera olika strategier för navigering. Arbetet har utförts utgående från en given plattform där fordonet endast använder sig av enkla ultraljudssensorer samt pulsgivare på hjulen för att mäta förflyttningar. Fordonet kan även autonomt navigera samt följa en enklare given bana i en känd omgivning. Systemet använder ett partikelfilter för att skatta fordonets tillstånd med hjälp av modeller för fordon och sensorer.</p><p>Arbetet är en fortsättning på projektet Collision Avoidance för autonomt fordon som genomfördes vid Linköpings universitet våren 2005.</p> / <p>In this thesis a system for navigation and control of an autonomous ground vehicle has been implemented. The purpose of this thesis is to further develop the vehicle that is to be used in studies and evaluations of path planning algorithms as well as studies of other autonomy functions. With different sensor configurations and sensor models it is also possible to evaluate different strategies for navigation. The work has been performed using a given platform which measures the vehicle’s movement using only simple ultrasonic sensors and pulse encoders. The vehicle is able to navigate autonomously and follow a simple path in a known environment. The state estimation is performed using a particle filter.</p><p>The work is a continuation of a previous project, Collision Avoidance för autonomt fordon, at Linköpings University in the spring of 2005.</p>

Mobile robots are commonly used to achieve tasks involving tracking a desired trajectory and following a predefined path in different types of terrains that have different surface characteristics. A mobile robot can perform the same navigation task task over different surfaces if the tracking performance and accuracy are not essential. However, if the tracking performance is the main objective, due to changing the characteristics of wheel-ground interaction, a single set of controller parameters or an equation of motion might be easily failing to guarantee a desired performance and accuracy. The interaction occurring between the wheels and ground can be integrated into the system model so that the performance of the mobile robot can be enhanced on various surfaces. This modeling approach related to wheel-ground interaction can also be incorporated into the motion controller. In this thesis study, modeling studies for a two wheeled differential drive mobile robot and a steerable four-wheeled robot vehicle are carried out. A strategy to achieve better tracking performance for a differential drive mobile robot is developed by introducing a procedure including the effects of external wheel forces / i.e, traction, rolling and lateral. A new methodology to represent the effects of lateral wheel force is proposed. An estimation procedure to estimate the parameters of external wheel forces is also introduced. Moreover, a modeling study that is related to show the effects of surface inclination on tracking performance is performed and the system model of the differential drive mobile robot is updated accordingly. In order to accomplish better trajectory tracking performance and accuracy for a steerable four-wheeled mobile robot, a modeling work that includes a desired trajectory generator and trajectory tracking controller is implemented. The slippage is defined via the slip velocities of steerable front and motorized rear wheels of the mobile robot. These slip velocities are obtained by using the proposed slippage estimation procedure. The estimated slippage information is then comprised into the system model so as to increase the performance and accuracy of the trajectory tracking tasks. All the modeling studies proposed in this study are tested by using simulations and verified on experimental platforms.

With the advent of faster computer processors and better optimization algorithms, Model Predictive Control (MPC) systems are more readily used for real-time applications. This research focuses on the application of MPC to trajectory generation of autonomous vehicles in an online manner. The operating environment is assumed to be unknown with various different types of obstacles. Models of simplified 2-D dynamics of the vehicle are developed, discretized and validated against a nonlinear CarSim vehicle model. The developed model is then used to predict future states of the vehicle. The relationship of the weight transfer to the tire slip angle is investigated. The optimal trajectory tracking problem is formulated in terms of a cost function minimization with constraints. Initially, a gradient descent method is used to minimize the cost function. A MATLAB based MPC controller is developed and interfaced with CarSim in order to test the controller on a vehicle operating in a realistic environment. The effects of varying MPC look-ahead horizon lengths on the computation time, simulation cost and the tracking performance are also investigated. Simulation results show that the new MPC controller provides satisfactory online obstacle avoidance and tracking performance. Also, a trajectory tracking criterion with goal point information is found to be superior to traditional trajectory tracking methods since they avoid causing the vehicle to retreat once a large obstacle is detected on the desired path. It is further demonstrated that at a controller frequency of 20Hz, the implementation is real-time implementable only at shorter horizon lengths. / UOIT