|dc.description.abstract||Multi-axle vehicles, such as trucks and buses, have been playing a vital role in trucking industry, public transportation system, and long-distance transport services. However, at the same time, statistics suggest more than one million lives are lost in road accidents each year over the world. The high adoption and utilization of multi-axle vehicles hold a significant portion of road accidents and death.
To improve the active safety of vehicles, active systems have been developed and commercialized over the last decades to augment the driver's actions. However, unlike two-axle vehicles (e.g., passenger cars), multi-axle vehicles come in a rich diversity and variety to meet with many different transportation needs. Specifically, vehicle configurations are seen in different numbers of axles, numbers of articulations, powertrain modes, and active actuation systems. In addition, multi-axle vehicles are usually articulated, which makes the dynamics and control more complex and challenging as more instability modes appear, such as, trailer sway and jackknife.
This research is hence motivated by an essential question: how can a universal and reconfigurable control system be developed for any multi-axle/articulated vehicle with any configuration? Leveraging the matrix approach and optimization-based techniques, this thesis developed a reconfigurable and universal modeling and control framework to this aim. Specifically, a general dynamics modeling that unifies any multi-axle and articulated vehicles in one formulation is developed in an intuitive manner. It defines the `Boolean Matrices' to determine any configuration of the articulation, the number of axles, and the active actuation systems. In this way, the corresponding dynamics model can be easily and quickly formulated when axles, articulations or actuators are added or removed.
The general modeling serves to achieve the universality and reconfigurability in controller design. Therefore, a hierarchical, i.e., two-layer, control system is proposed. In the high layer, the optimization process of a model predictive control (MPC) calculates corrective Center of Gravity (CG) forces/moments, which are universal to any vehicle. The lower-level controller is achieved by a Control Allocation (CA) algorithm. It aims to realize the MPC commands by regulating the steering or torque (driving or braking) at each wheel optimally. In addition, the optimization takes into account real-time constraints, such as actuator limits, tire capacity, wheel slips, and actuators failure.
Simulations are conducted on different vehicle configurations to evaluate control performance, reconfigurability, and robustness of the system. Additionally, to evaluate the real-time performance of the developed controller, experimental validation is carried out on an articulated vehicle with multiple configurations of differential braking systems. It is observed that the controller is very effective in dynamics control and has a promising reconfigurability when moving from one configuration to another.||en