The increasing number of active chassis systems leads to a higher need for control integration: coordinating the vehicle actuators in a supervised way has the potential of enhancing the vehicle performance in terms of handling and power efficiency while addressing the possible conflicts between the different actuators, thus reducing implementation cost of active systems thanks to better reusability, ease of configuration and calibration. This research presents the development of a multi-layered control framework for vehicles equipped with an electric drivetrain, independent braking, and active steering actuators. The control architecture is decomposed into two parts: an abstract layer that defines controls at the wheel level and which does not require precise knowledge of the actuators equipped on the vehicle, and an application layer that coordinates the actuators to follow the wheel control requests. A simplified tire model is developed to model the coupling behavior between the longitudinal and lateral tire forces; the abstract layer utilizes multivariable control methods in conjunction with the simplified tire model to define optimal wheel controls. Similarly, a control allocation is implemented in the application layer to coordinate the brake and drivetrain actuators. The distribution of actuator commands is made invisible at the wheel torque level by cleverly using the Smith-McMillan decomposition of a redundant system, simplifying the controller design by dissociating the actuator allocation problem from the control problem while ensuring internal stability and good robustness properties. Simulations with a high-fidelity vehicle model validate the control framework. Several actuator configurations are considered to highlight the reusability of the control architecture. Results show that the control architecture provides a unified framework for the vehicle's longitudinal, lateral, and yaw control.

The automotive industry is in the midst of a shift towards electrification, with an increasing prominence of by-wire technologies. This thesis is dedicated to investigating the modeling, control, and parameter optimization of brake-by-wire actuators, with applications for both future autonomous and non-autonomous vehicles. To this aim, a modular control architecture is proposed to facilitate the integration of these actuators into vehicles.

The actuators are initially modeled through bond graph methodology. Subsequently, a cascaded control approach is employed to govern the intelligent actuators, with individual controllers designed using the Youla parameterization technique. Two distinct physical parameter optimization strategies are employed to enhance actuator responsiveness and energy efficiency. One approach is based on linear system optimization using transfer function, while the other centers around nonlinear system optimization. Comparative evaluations of the actuators are conducted using step and ramp tests. Finally, a direct comparison of the actuators within the proposed control architecture is carried out through a straight-line braking test scenario.

This thesis will present the development and construction of a scale autonomous vehicle, along with itssoftware development. The platform will serve as a means to test various lateral controllers and compare their results. Using Buckingham Pi Theorem, the platform will be scaled to achieve dynamic similitude with a standard electric vehicle, while utilizing common sensors found in modern autonomous vehicle systems. The software framework will be created using modern software packages, allowing for easy integration with a wide range of useful robotics tool kits and providing the ability to modify the car for various future applications. The study will examine the performance of a variety of control methodologies, with consideration for both their simulated and actual performance on the scaled car.

Adverse weather conditions, mainly uncertain road surface conditions, are a remaining challenge in the field of lateral motion control of an Automated Driving System. This thesis proposes treating this challenge as one of addressing model parameter uncertainty. This perspective is framed as a multi-objective optimal robust control problem. The first objective is to simultaneously minimize yaw and lateral position reference tracking errors despite the condition of the road surface. The second objective is to minimize actuator effort. To ensure that the solutions remain practical, this thesis constrains the possible solutions to \textit{cheap} control techniques (techniques that use low amounts of runtime and memory).

Before developing controllers, this thesis first develops a new methodology that can be used to determine the amount of tracking error that can be tolerated to achieve an overall target level of safety for the entire Automated Driving System. The new methodology combines three techniques: (1) risk allocation, (2) geometric analysis of the driving task, and (3) statistics. The methodology argues that the control requirements must be designed with consideration of the performance of upstream modules such as planning and localization.

To investigate and compare the current state-of-the-art in this field, a high-fidelity simulator is developed with the commercial software CarSim. This simulator enables the simulation of many lateral motion controllers in a wide variety of maneuvers and environmental conditions. Following the development of this simulator, the bicycle car model is used to develop different approaches to lateral motion control. This analysis presents new understandings of how the controller's performance is influenced by the many possible definitions of the system's state. Lateral motion control is then separated into trajectory tracking and path tracking. For the remainder of the thesis, path-tracking is used because it is shown to better decouple lateral motion control performance from longitudinal motion control.

One of the challenges of designing a robust controller is balancing conservatism with performance. To address this, a new Linear Parameter Varying (LPV) control technique is developed. This new LPV control technique allows for the design of a controller whose parameters are continuously varying with the scheduled parameters without requiring special characteristics of the model (such as being affine with the scheduling parameters). This approach is used to design a high-performance controller in simulation. Real-world vehicle results on closed-course maneuvers and public road routes show that this new design performs well. The most significant result is that this new LPV control achieves a sufficient balance between tracking performance and passenger comfort for the velocity range of 5-30 m/s on concrete road surfaces and for highway-like and collision-avoidance maneuvers. This same controller also achieves this satisfactory performance on a gravel road between 5 and 15 m/s.

The next investigation compares the performance of several controllers that require low computational resources. These results, collected in simulation, argue that many path-tracking control problems can be solved by cheap controllers. More complex controllers are not necessary and instead the researchers in this field should focus less on the commonly solved problems such as lane-keeping, lane-changes, collision-avoidance and more on limit handling, adverse road surface conditions, fault tolerance, and other practical challenges.

The final investigation shows that steering dynamics (which are often neglected in lateral motion control) can play a significant role in performance and need to be included in the overall control design. In this investigation, the problem of having little information on the steering actuator is addressed by developing a new data-driven model that captures key characteristics of the actuator. The first characteristic accurately modeled is the dependence of the static gain on velocity. The second characteristic modeled is the decrease in system damping as velocity increases. A steering controller is then developed using this new model. An outer loop controller performs path-tracking control. On-vehicle results show that the resulting design works well on highways in ideal and rainy conditions.

Heavy-duty electric and hybrid electric vehicles are potential means to reduce the emissions of the transportation sector. However, the lithium batteries needed to power these vehicles can be cost and weight prohibitive, and battery degradation adds to the lifetime cost of these vehicles. Buses in particular are considered throughout this work—their frequent stopping and starting makes them prime candidates for electrification or hybridization, yet that same stopping and starting can be a source of significant battery wear. This work explores methods to improve battery lifespan and improve the overall economic feasibility of heavy-duty alternative powertrain vehicles.Four studies are carried out to this effect; simulation is used in all cases due to the slow rate of battery degradation and the expense associated with destructive testing. First, an electric bus is fitted with an on-board photovoltaic system. A full model for on-board photovoltaics is developed and it is shown that the power provided by the modules reduce the battery discharge depth to a sufficient degree to improve battery lifespan. Bus rooftop photovoltaics are shown to have a positive return on investment. Next, aging-aware control of a hybrid energy storage system is considered. Hybrid energy storage pairs an ultracapacitor with the conventional lithium battery to reduce large current spike and improve battery aging. A new energy management strategy that incorporates ultracapacitor aging is shown to be a more effective means of control than existing literature. The third and fourth studies concern robust energy management. The third considers the robustness of aging-aware energy management to aging model variations and methods of improving the robustness of aging-aware strategies are proposed. The fourth study introduces a new energy management concept that incorporates elements of minimax dynamic programming. This new strategy is first shown to improve robustness of a series hybrid bus to driving condition uncertainty, then second it is shown to improve the performance of aging-aware control of an electric vehicle with hybrid energy storage.