Graduation Year

2021

Document Type

Dissertation

Degree

Ph.D.

Degree Name

Doctor of Philosophy (Ph.D.)

Degree Granting Department

Mechanical Engineering

Major Professor

Tansel Yucelen, Ph.D.

Committee Member

Jonathan A. Muse, Ph.D.

Committee Member

Rajiv Dubey, Ph.D.

Committee Member

Kyle Reed, Ph.D.

Committee Member

Yasin Yilmaz, Ph.D.

Keywords

Uncertain Dynamical Systems, Distributed Control, Event-triggering, Model Reference Adaptive Control

Abstract

Despite the numerous advantages that networked controlled systems provide, potential network overload in these systems can severely degrade their performance and even cause catastrophic failures. To this end, the work presented in this dissertation provides a neat solution on reducing network utilization in uncertain networked controlled systems using an adaptive event-triggered control approach predicated on a novel dynamical threshold and information exchange methodology while guaranteeing closed-loop system stability.

In particular, consider a system consisting of an operator (e.g, a ground station) and an uncertain controlled dynamical system (e.g., a vehicle equipped with feedback control algorithm), where the aim of the operator is to provide guidance commands (e.g., trajectory information) to the controlled dynamical system through a wireless network. To reduce wireless network utilization between the operator and the controlled dynamical system, we propose a new event-triggering approach for scheduling networked guidance data transmissions. A reference model of the dynamical system is available to and utilized by the operator such that we explore three event-triggering rules based on an error signal. This error signal is a function of the reference model states and the guidance commands. The first event-triggering rule is the standard one in the literature where sampled data points of the guidance commands are exchanged between the operator and the controlled dynamical system when an event occurs. The second and third rule deviate from the standard one in the literature and are respectively predicated on exchanging approximated curve-fitted guidance command functions estimated using the current guidance command value and future waypoints, and exact guidance command functions valid over a time-interval. We further build on this methodology for scheduling networked guidance data to encompass a class of uncertain dynamical systems by augmenting the nominal feedback control algorithm with an adaptive counterpart. To elucidate the efficacy of the proposed event-triggering control approach for networked systems we provide an experimental study using crazyflies nano-quadcopter from bitcraze as a vehicle equipped with feedback control algorithm and a PC running python codes as an operator. In addition, we provide a numerical study of an inverted pendulum model as an uncertain dynamical system to show that the the second and third event-triggering rules further reduce network utilization.

In contrast to their continuous-time counterpart, discrete-time networked control systems have a built-in feature of reducing wireless network utilization between an operator (e.g., ground station) and a controlled dynamical system owing to their intrinsic property of sampling the data at a chosen rate. Furthermore, discrete-time control strategies are known to surpass their continuous-time counterpart in execution complexity, since they can be directly executed in embedded code. Motivated by these advantages, we extend our event-triggering approach to discrete-time systems and provide stability analysis accompanied with numerical study.

Going further, multiagent systems are a big and important subset of networked controlled systems. In their design and implementation, in addition to guaranteeing closed-loop system stability, it is essential to provide a scheme for scheduling inter-agent information exchange to prevent potential network overload and decrease wireless communication costs. To this end, we propose a new event-triggered distributed control architecture. Yet again, owing to its efficacy, we develop a new dynamical threshold predicated on an error signal between the system states and its reference model states and an exponentially decaying term. Moreover, we introduce a method called solution-predictor curve exchange, where this method is capable of predicting the time-solution trajectory related to information exchange such that each agent stores this curve and distributively exchanges its parameters when an event occurs. The key feature of this method is that each agent utilizes the resulting solution trajectory over the time interval until the next event occurs, where it has the capability to further reduce inter-agent information exchange as compared with the sampled data case. We provide rigorous system-theoretical analysis capturing both, sampled-data exchange and solution-predictor curve exchange. Through a numerical study we provide practical guidelines for selection of event-triggering parameters. Furthermore, elucidate the efficacy of the proposed distributed event-triggered control architecture predicated on the solution-predictor curve through an experimental study using crazyflies nano-quadcopters from bitcraze.

As the behavior of many dynamical systems can be closely captured using linear time-invariant models, we extend our distributed event-triggered control architecture form single-integrator models to linear time-invariant models. Our first proposed approach requires that agents exchange both, their output values and the output values of their corresponding reference models. We then propose an event-triggered control architecture for linear time-invariant systems that removes the requirement of exchanging the output values of the reference model. Finally, we propose distributed adaptive event-triggered control architecture to capture a class of uncertain dynamical systems. Again, in addition to the rigorous system-theoretical analysis, we show the efficacy of the proposed distributed event-triggered architecture predicated on the solution-predictor curve exchange through an illustrative numerical study.

The presence of actuator dynamics in the system can significantly degrade its performance and even lead to a failure. Even though there are extensive studies in the literature to deal with the presence of actuator dynamics with full state feedback, how this extends to output feedback is unknown. We propose an output feedback control architecture to breach this knowledge gap. In addition, it is well known that adaptive control impairs transient performance and can require extensive computational resources when the state of the dynamical system is large. To this end, we propose a reduced-order model reference adaptive control for transient performance improvement.

Share

COinS