The world is becoming increasingly congested due to a continuous growth of world population and overall wealth. Current road capacity limitations will lead to a significant decrease of traffic flow accompanied by a serious increase in global fuel consumption and air pollution. Ever-expanding road infrastructure is not expected to be a sustainable and long-term solution, thus other solutions are sought. Amongst these solutions is the concept letting AVs drive closer together, cooperatively. A flock of vehicles that coordinates collective movement while maintaining short inter-vehicle distances is defined as platooning. Evidently, in order to achieve safe platooning, a thorough understanding of both longitudinal and lateral behavior of the platoon system and its dynamics is required. The longitudinal aspect of platooning concerned with the design of spacing policies (i.e., distance keeping) has been broadly researched in the past decades. As a consequence, numerous valid applications exist. Whereas for the lateral aspect, the subject has not been researched as extensive. Hence a considerable amount of knowledge on this side is still needed to meet strict conditions and requirements for platooning applications. One of the major bottlenecks obstructing robust lateral platoon control is the ability to assure the lateral string stability for the complete platoon. String stability implies that errors propagating in an upstream direction of inter-connected vehicles forming the vehicular platoon, do not amplify. Specifically, lateral string stability implies that initially bounded lateral errors will remain bounded between ever pair of consecutive vehicles along the string of vehicles. Propagating errors are therefore attenuated during, e.g. the execution of a lane change. Henceforth, when Lateral String Stability for a platoon can be guaranteed, so is the reassurance that a platoon can safely perform certain manoeuvres such as a collective lane change. This thesis endeavours to develop a string stable, lateral controller for a homogeneous platoon of vehicles using a Model Predictive Control (MPC) approach. In the process two different control strategies tightly linked with the information flow topology, being centralized and distributed, are designed and compared in terms of reference tracking performance, noise- and disturbance rejection and practical implementation. Lastly, the developed controllers are simulated and validated using Siemens' Simcenter Prescan software, after which the results are thoroughly discussed. Results have indicated that for the application discussed in this work, the centralized controller outperformed its competitor in the field of tracking performance and noise rejection, but not by a great margin. Furthermore, the novel developed definition of Practical Lateral String Stability (PLSS) guarantees stability for a platoon of n=5 vehicles while using both controllers. To this end, the distributed controller is seen as worthy competitor and more workable solution due to the centralized controller's issue of practical implementation. As part of anticipated future work, we plan testing the proposed approach with field experiments to validate the proposed method in real life.

Last decades the autonomous driving research field has shown exponential growth. The social benefits, which include increased safety, mobility and productivity, are the main factor that drive this growth. One of the most difficult problems that vehicle engineers must solve to develop autonomous vehicles is the motion planning problem. They must solve the motion planning problem for environments ranging from unstructured to structured, such as parallel parking up to high-speed highway driving. Current literature presents many implementations that solve either the structured or unstructured planning environment or a small range of environments. Yet, the generic implementation of a single motion planning method, that can plan in the full range of environments, is still an open question. The aim of this thesis is to address the identified gap in the literature, by realizing a real-time implementation of a single motion planning method, that shows human-like and safe driving behavior, and can deal with any environment it encounters.
In this thesis, a method is proposed that solves the planning problem by enhancing the Closed-Loop Rapidly-exploring Random Tree (CL-RRT) algorithm for planning on curved structured roads. The planner is aware of the road curvature and deforms the motion plan, so it follows the shape of the road. Extensive simulations have demonstrated that the proposed method can improve the path quality on curved highway roads when compared with the standard RRT and CL-RRT. Although the method can plan in any environment it encounters, it demonstrated limitations in its capability of dealing with complex dynamic environments.

In the automotive industry automation is popular and every year car OEMs advance their technology to be able to drive autonomously. Longitudinal control of the vehicles is an important part of the complete autonomous driving system. The difficulty of this control problem lies with changing longitudinal dynamics and the lack of full-state system information. This complicates controller design when using classic model-based approaches such as Optimal Control (OC). Currently the controllers are still manually tuned by control engineers in the
vehicle. This is time consuming and expensive, therefore other methods for controller design such as learning are explored. Reinforcement Learning (RL) is one of those methods. To examine the potential benefits of learning a controller, this work will make a comparison between RL and OC. For RL, an actor-critic structure using deterministic policy gradient is applied. Due to partially observable system dynamics OC is used as an optimal output feedback controller. The comparison complies speed control of an autonomous vehicle. The RL agent will learn a controller by training on a nonlinear high fidelity vehicle model. In this work it was demonstrated that RL can reach the same performance as OC when all environmental settings are comparable. When environmental settings deviate, it was is found that RL outperforms OC. To verify the simulated results all controllers were confirmed in an experimental real-life setting.In conclusion, this proved a promising benefit of learning with respect to classical controller computation, when dealing with partially available system information.

In modern society cars are one of the most important means of transportation. Unfortunately, many people die in car accidents around the world. Research shows that the number of fatal casualties in car accidents has been increasing for the past decade and that the largest cause of these accidents is the human driver. For this reason, research on fully autonomous vehicles has gained a lot of attention. However, currently autonomous driving is only implemented to reduce the errors of human drivers. More research is necessary in order for fully autonomous vehicles to be implemented and to remove the human driver completely. A robust navigation algorithm which is able to run in real time is one of the challenges in development of fully autonomous vehicles. Important topics in navigation of autonomous vehicles include the path planner and the motion controller. The path planner finds a path for the vehicle from its current location to the target location. At the same time the path planner avoids obstacles and fulfills the non-holonomic constraints of the autonomous vehicle. The motion controller tries to follow the path the path planner made as close as possible by controlling the vehicle. These two topics influence each other and are therefore dependent. In literature little research is done on integrated algorithms that combine path planning and motion control. Therefore, this thesis will research navigation of autonomous vehicles by using an integrated algorithm that includes both path planning and motion control. The objective of this thesis is to develop a Lyapunov stable control algorithm that is capable of planning a path for all possible vehicle maneuvers. Besides path planning the proposed algorithm must be capable of controlling the vehicle along this path. Furthermore, the algorithm needs to include obstacles and the non-holonomic dynamics of an autonomous vehicle. The main contribution of this thesis is an integrated path planner and motion controller for navigation of autonomous vehicles. The stability of the proposed algorithm is proven by using the Lyapunov method. Simulation results prove that the algorithm is capable of planning the path and the motion of the autonomous vehicle with non-holonomic constraints and with the presence of obstacles.

The work presented in this report focuses on longitudinal control of heterogeneous platoon. Firstly, merging maneuver is looked upon, considering heterogeneity in the vehicles i.e. considering different dynamics for the vehicles involved. Then heterogeneity in platoon, arising from the difference in potential limits of the vehicles is taken into account, which the current literature fails to address. The uncertain/heterogeneous dynamics are addressed through the difference in the engine driveline constant, which differs for different vehicles and keeps on changing during operation. The heterogeneity is tackled by using Adaptive control, which adapts to the changing dynamics and estimates the uncertainty to act accordingly. Adaptive strategies for formation keeping in platoons of automated vehicles have been recently proposed; while these strategies are able to cope with uncertain vehicle parameters, they have the drawback of handling only acyclic graphs (e.g. look-ahead topology). This prevents from enhancing formation keeping protocols with more complex platooning maneuvers such as synchronized merging/splitting. This work proposes an adaptive strategy for performing synchronized merging maneuvers in the presence of uncertain vehicle parameters: during these maneuvers, a cyclic communication graph is instantiated, which must be handled in a suitable way. The strategy is framed as a synchronization protocol with a set of adaptive control laws, designed via Lyapunov stability theory. A benchmark scenario in which two platoons formed in different lanes are required to merge, (e.g. due to a lane closure because of roadworks) is presented to show the effectiveness of the proposed strategy. The approach is proven to be scalable to platoons with arbitrary number of vehicles. In later part of the work, a method is proposed which takes the limits of the vehicles into consideration, to make the platooning realistic. An adaptive platooning strategy is used so as the vehicles can adaptively synchronize to desired reference dynamics. In order to handle engine limits, a mechanism is proposed based on making the reference dynamics `not too demanding', by properly saturating their control action. Such saturation action will allow all vehicles in the platoon to remain in their engine potential limits throughout. Another problem that affects the stability of the platoon is that, any perturbations or disturbance in the platoon leaves the vehicle scattered if the vehicles are already running at their potential limits. This work proposes an adaptive platooning method with bi-directional interaction and a mechanism coping with engine constraints. The bi-directional interaction allows the vehicle to not only look at the error in spacing from the preceding vehicle (vehicle in front) but also the succeeding vehicle, this makes the vehicle's control action (acceleration) depend on both front and back errors, the vehicle can thus slow down for the following vehicle if it lags behind. The bi-directional strategy developed is then proved string stable, which the literature struggles to establish with the already developed bi-directional strategies. Simulations are then conducted to validate the theoretical analysis and show the effectiveness of the method in retaining cohesiveness of the platoon. The simulations include a merging maneuver to check the effectiveness of merging with synchronized cyclic communication. For the engine potential limits, the firstly state of the art algorithm's simulation is run which is then compared to the performance of the proposed algorithm. Finally, to check the effectiveness of bi-directional CACC strategy simulations for a specific case is run for both uni-directional and bi-directional strategies proposed in the thesis and are then compared, showing the advantage of using bi-directional CACC strategy over uni-directional.

This report presents the results of the graduation thesis from the TU Delft, performed at the Integrated Vehicle Safety department of TNO. The goal of this graduation project is to develop a unified path planning and tracking method for autonomous vehicles in highway driving scenarios, by making use of Artificial Potential Fields (APFs).

Most autonomous vehicles base their navigation control on first planning a path, which is then tracked by using a combination of feedback and feedforward control. The strength of using APFs is the possibility to integrate the path planning and tracking process. This concept has been extensively used in the field of robotics. The attractive and repulsive forces coming from the APF guide the robot towards the final goal while avoiding obstacles. This offers an intuitive way to represent the level of hazard experienced in the direct environment. However, considerably less research has been devoted to the application of APFs in the field of autonomous vehicles. Furthermore, the available research mostly treats the vehicle as a particle, thereby leaving out the more complicated vehicle dynamics.

Therefore, this research is aimed at including the vehicle dynamics into the path-planning process such as to generate feasible and desirable paths. A Model Predictive Control (MPC) framework is proposed to fulfill this task. The adopted vehicle model is given by the linear bicycle model, which represents the vehicle dynamics sufficiently well for highway applications. The two main manoeuvers of lane keeping and lane changing are executed with the aid of two different APFs that were designed for these specific purposes, respectively. A second order Taylor approximation is used to incorporate the APFs into the quadratic MPC cost function.

The Simulink model from TNO to simulate the controlled vehicle is modified by extending it to curved roads and by including the developed APF MPC algorithm. The resulting algorithm is capable of following curved highway lanes and overtaking slower vehicles at different velocities in the simulation environment. The results are compared with the previously developed path planner and lateral controller from TNO. In order to also deliver longitudinal control action, the model can be connected to the adaptive cruise control application of TNO.

It is concluded that the suggested method potentially offers a powerful solution to the navigation control of autonomous vehicles. Real-life experiments have to be done in the future to validate the performance of the controller.

Over the past years, the automotive industry has seen a constantly increasing level of automation of automotive vehicles. This increasing level of automation contributes to an increasing safety in traffic and a reduction of traffic congestions due tio faster response times and higher reliability, with respect to the human driver. The automation of a vehicle over the longitudinal and lateral degree of freedom requires additional safety measures to ensure safety of the passengers during automated maneuvers. In a lateral control scenario, the fault detection and isolation of faults occuring in the steering system belongs to these set of safety measures. By detecting and isolating faults of interest in this system, a decision process follows which will, for example, either compensate for the acting fault or bring the vehicle to a safe standstil in the case that the fault surpasses a non-acceptable threshold.

This work presents a set of novel methods for fault detection and isolation for a generalized set of linear time-invariant and parameter-varying systems. The faults under investigation comprise of an additive fault acting as an offset on the system and a multiplicative fault acting non-linearly on a set of known signals. Furthermore, the system is subjected to exogenous disturbances. The first challenge imposed is the lack of isolability of the additive and multiplicative fault using conventional linear estimation techniques. The second challenge imposed is the design of a fault detection mechanism for linear parameter-varying systems subject to exogenous faults and disturbances.

The first contribution of this research is a moving least-squares based approach as an extension to an existing nullspace computation based parity-space fault detection filter. This novel combination allows a decoupled estimation of the additive and multiplicative fault. The second contribution is an extension to the first contribution by attacking one of the largest sources of error: the dynamical content of the parity-space filter. The estimation performance of the combined fault detection and isolation filter of both contributions is provided with a guaranteed performance bound, providing not only an intuitive tool to push down the estimation error, but also shows possible trajectories for future work. The third and final theoretical contribution is the adaption of the linear time-invariant parity space method for a linear parameter-varying environment. A convex quadratic optimization problem is used which has an exact analytical solution for approximately rejecting the parameter-varying effects of the system.

A contribution of the more practical aspect of this thesis, is a demonstration of the applicability of the developed framework on real-life problems, both in a simulation setting and an experimental setting. The theoretical framework has been applied in a practical case study in fault detection and isolation for lateral control of autonomous vehicles. In this application, there is a particular interest in detecting and isolating the additive and multiplicative fault, acting on the steering mechanism of the vehicle, for compensation or decision making of the autonomous system in faulty scenarios. The experimental results confirm the theorems, showing that the faults can be detected and isolated both in theory and in practice.

The highly competitive nature of Formula one makes every team eager to find areas of improvement on their cars. In the case of Scuderia Toro Rosso, it was found that there is a gap to improve performance while driving on kerb stones. Analysis of competitors showed that they can take wider driving lines with more speed. In order to improve this performance, the need for a high fidelity dynamic car model arose. This model can be used to understand the car dynamics, which can be used to choose a better suspension setup.

The baseline model used for this study was developed in MSC.Adams. After a discussion with field experts, two areas of model improvement were identified. In the baseline model, the tire is modeled using a MF-Tire model, where the vertical dynamics was modeled as a single spring and damper. In order to simulate the tire dynamics on kerb stones, a more advanced model is necessary. For this research, FTire was chosen as the best option. The other area of improvement is modeling the compliance of the suspension system. In the baseline model,
all suspension members are modeled as rigid links, without any compliance. To model this compliance, an ’equivalent stiffness’ model was proposed, where all stiffnesses are combined into flexible joints.

To use the FTire model, a set of parameters needs to be found for the Formula one tires. To find these parameters, measurements were done to capture the dynamics of the tire. A method is developed to filter out any rig disturbance, and make the data useful for parameter estimation. The estimation was done using Matlab, with MSC.Adams and FTire in-the-loop, and made use of a Genetic Algorithm (GA). Comparing the simulated behavior with the measurement showed good correlation for various conditions.

Since every part in the car contains some compliance, modeling and measuring every single part would be an extensive task. In order to reduce the complexity, all compliances are combined into a simplified model. In this simplified model, all compliances are modeled in the joints. A Pattern-Search algorithm in Matlab is used in combination with MSC.Adams to find the joint stiffnesses in such way the model matches measurements of the overall car compliance.

To measure the overall improvement, both models are combined into a full car model. The baseline and proposed model are validated against measured track data. From this compare, it can be concluded that the proposed model shows a significant better correlation on kerb simulations. This is mainly due to the fact that the baseline tire model is not capable of simulating the harsh road input. The proposed model provides a more powerful tool for engineers to better understand the car dynamics, and to find the best suspension setup for
kerb riding.

Over the past century, cars have become the most important means of transportation. With the increasing number of vehicles on the road, safety has become an important topic. The introduction of driver assistance systems, like Electronic Stability Control, has led to a significant decrease in fatal accidents. By preventing the vehicle from entering unstable behaviour, it remains controllable for the average driver. In rally racing, though, unstable manoeuvres, like drifting, are widely used for improved agility on low friction surfaces and in sharp corners. The average driver, however, is unable to perform such a manoeuvre safely.

With the development of autonomous driving systems, steps are taken to increase driver safety by removing the effects of human error. By taking over full control of the vehicle, these systems do not need to consider the driving skills of the average driver. Furthermore, the driving skills of the autonomous driving system are a result of its design. The autonomous driving system could, therefore, be designed to use advanced driver techniques, like drifting, to increase the safety of the passengers and other road users.

The objective of this research is to develop a vehicle controller that is capable of path-tracking within and beyond the limits of stable handling. To reduce the costs of testing, the controller is developed for an experimental platform in the form of a 1/10 scale radio controlled car. A nonlinear vehicle model of the scaled vehicle is developed and analysed regarding unstable vehicle behaviour. Since the behaviour and actuation of a vehicle differ in typical cornering and limit handling conditions, two separate controllers are used. In typical cornering conditions, a steering controller is used for path-tracking. Beyond stable limit handling, the lateral motion is actuated by both the steering angle and throttle input. A drift controller is used to determine the appropriate actuator inputs to sustain the drift, based on the yaw rate and direction of motion. The main contribution of this research is an extension of that controller to add path-tracking capabilities. Simulation analysis shows the controller can enter and sustain a drift, while successfully tracking the given path. Implementation of the controller in the scaled car shows great path-tracking performance in typical cornering conditions. Furthermore, the controller is able to, on command, make the vehicle enter sustain a drift.

Automated vehicle control provides advantages in transport efficiency, redundancy, human-safety, flexibility, and parallelism. Extensive research has been dedicated to direct-following lateral control methods, using a single preview point as reference signal. However, these methods give rise to significant tracking errors. As alternative to a single point, a continuous path can be used as reference signal for vehicle control. Using a continuous reference path, accurate and comfortable control actions can be computed, while avoiding reference tracking errors. Therefore, robust reference path generation is an essential part of lateral vehicle control.

The goal of this research is to develop a generic, robust reference path generator for lateral vehicle control. Currently in path generation, the state-of-the-art method is based on repetitive polynomial fitting. This method inherently contains two main weaknesses. Firstly, it is not robust to sensor noise and other real-world disturbances. Secondly, as a result of the repetitive fitting, a discontinuous path is generated. This is undesirable, because it leads to an unfeasible reference path, since vehicles can only produce and track continuous trajectories. Moreover, a discontinuous reference path results in the need for path smoothing when used in comfortable lateral vehicle control. Fundamentally, this smoothing leads to inaccurate control, caused by manipulation of the original, discontinuous reference path. To overcome these weaknesses, a new Model-based Path Generation method is presented.

The development of a new, generic method for robust path generation is the main contribution of this research. This new path generation method is capable of producing feasible vehicle trajectories based on unfeasible waypoints. The performance of this method is evaluated in simulations and experiments, benchmarking it against polynomial fitting path generation. Furthermore, path generation robustness is assessed based on the outcome of a disturbance sensitivity analysis. The results are in accordance with the hypothesis stating that the new method outperforms the benchmark method in terms of path accuracy, robustness, continuity, and general applicability.

This MSc. thesis explores the design and implementation of a motion planner for non-holonomic autonomous vehicles in parking spaces. The planner must avoid collisions with static obstacles, satisfy performance, comfort and safety constraints for the motion, satisfy the non-holonomic constraints of the vehicle model, consider the dynamics of the vehicle actuators and be fast enough for real-time implementation.

The motion is planned by solving an Optimal Control Problem (OCP) that is discretized in time in order to obtain a non-linear, non-convex, multi-variable optimization problem. The thesis addresses how to solve this optimization problem so that it results in motions that satisfy the requirements. Specifically, the motion is split into a number of waypoint tracking sections, planned by correspondingly simplified optimization problems. The thesis also addresses methods to further simplify the optimization process in order to reduce the implementation time.

The results show the planner satisfies the constraints but does not quite work in real time. Recommendations for improving the planner in the future are given in the report.