<div>In recent years, the collective behaviors in nature have motivated rapidly expanding research efforts in the control of multi-agent systems. A multi-agent system is composed of multiple interacting subsystems (agents). In order to seek approaches that respect the network nature of multi-agent systems, distributed algorithms has recently received a significant amount of research attention, the goal of which is allowing multi-agent systems to accomplish global objectives through only local coordination. </div><div> Under this scope, we consider three major problems in this dissertation, namely, distributed computation, distributed optimization, and the resilience of distributed algorithms. First, for distributed computation, we devise distributed algorithms for solving linear equations, which can eliminate the initialization step for agents; converge to the minimum $l_1$ and $l_2$ solutions of under-determined linear equations; achieve ultimate scalability inters of agents' local storage and local states. Second, for distributed optimization, we introduce a new method for algorithm discretization so that the agents no longer have to carefully choose their step-size. We also introduce a new distributed optimization approach that can achieve better convergence rate with lower bandwidth requirement. Finally, for the resilience of distributed algorithms, we propose a new approach that allow normal agents in the multi-agent system to automatically isolate any false information from malicious agents without identification process. Though out the dissertation, all mentioned results are theoretically guaranteed and numerically validated.</div>

Industry 4.0 holds many promises for manufacturers; however, a shortage of qualified employees has prevented a swift adoption of the revolution's new technologies. Engineer and Economist Klaus Schwab argues Education 4.0 is the key to addressing the employee shortage and preparing future generations for the shifting labor market. To support Education 4.0, classes must allow students to engage emerging technologies that help bridge Operational Technology (OT) and Informational Technology (IT). The thesis detailed an educational laboratory that demonstrates the application of data analytics (an IT tool) and optimize the performance of a cyber-physical system composed of industrial (OT) components. The lab experience focuses on a disc's controlled positioning (levitating) using a PLC-based PID controller and a VFD. The activity requires students to capture data of a moving discs, create a machine learning function representing the disc's movement, and use the machine learning function for classification and PID optimization problems. A comparative analysis of a PID cycle ensures a regressions model accurately represents the physical model using measurements including peak-overshoot, rise time, settling time, and the flight plots' Means of their Squared Error. Further, the study examines multiple ML models each built using various features to identify the systems relevant and redundant data.<br>

<div>Reinforcement Learning (RL) is a data-driven technique which is finding increasing application in the development of controllers for sequential decision making problems. Their wide adoption can be attributed to the fact that the development of these controllers is independent of the</div><div>knowledge of the system and thus can be used even when the environment dynamics are unknown. Model-Based RL controllers explicitly model the system dynamics from the observed (training) data using a function approximator, followed by using a path planning algorithm to obtain the optimal control sequence. While these controllers have been proven to be successful in simulations, lack of strong safety guarantees in the presence of noise makes them ill-posed for deployment on hardware, specially in safety critical systems. The proposed work aims at bridging this gap by providing a verification framework to evaluate the safety guarantees for a Model-Based RL controller. Our method builds upon reachability analysis to determine if there is any action which can drive the system into a constrained (unsafe) region. Consequently, our method can provide a binary yes or no answer to whether all the initial set of states are (un)safe to propagate trajectories from in the presence of some bounded noise.</div>

<div>The focus of this Dissertation is to provide a unified and efficient solution method for an important class of nonconvex, nonsmooth, constrained optimization problems. Specifically, we are interested in problems where the objective function can be written as the sum of a smooth, nonconvex term, plus a convex, but possibly nonsmooth, regularizer. It is also considered the presence of nonconvex constraints. This kind of structure arises in many large-scale applications, as diverse as information processing, genomics, machine learning, or imaging reconstruction.</div><div></div><div>We design the first parallel, asynchronous, algorithmic framework with convergence guarantees to stationary points of the class of problems under exam. The method we propose is based on Successive Convex Approximation techniques; it can be implemented with both fixed and diminishing stepsizes; and enjoys sublinear convergence rate in the general nonconvex case, and linear convergence case under strong convexity or under less stringent standard error bound conditions.The algorithmic framework we propose is very abstract and general and can be applied to different computing architectures (e.g., message-passing systems, cluster of computers, shared-memory environments), always converging under the same set of assumptions. </div><div></div><div>In the last Chapter we consider the case of distributed multi-agent systems. Indeed, in many practical applications the objective function has a favorable separable structure. In this case, we generalize our framework to take into consideration the presence of different agents, where each one of them knows only a portion of the overall function, which they want cooperatively to minimize. The result is the first fully decentralized asynchronous method for the setting described above. The proposed method achieve sublinear convergence rate in the general case, and linear convergence rate under standard error bound conditions.</div><div></div><div>Extensive simulation results on problems of practical interest (MRI reconstruction, LASSO, matrix completion) show that the proposed methods compare favorably to state-of-the art-schemes.</div>

Spacecraft formation flying refers to the coordinated operation of a group of spacecraft with a common objective. While the concept has been in existence for a long time, practical fruition of the ideas was not possible earlier due to technological limitations. The topic has received widespread attention in the last decade, with the development of autonomous control, improved computational facilities and better communication technology. It allows a number of small, lightweight, economical spacecraft to work together to execute the function of a larger, heavier, more complex and expensive spacecraft. The primary advantage of such systems is that they are flexible, modular, and cost-effective.<div><br></div><div>The flexibility of formation flying and other derived concepts comes from the fact that the units are not physically attached, allowing them to change position or orientation when the need arises. To fully realize this possibility, it is important to develop methods for spatial reorganization. This thesis is an attempt to contribute to this development. </div><div><br></div><div>In this thesis, the reconfiguration problem has been formulated as a single system with multiple inputs and multiple outputs, while preserving the individuality of the agents to a certain degree. The agents are able to communicate with their neighbors by sharing information. In this framework, a distributed closed-loop stabilizing controller has been developed, that would drive the spacecraft formation to a target shape. An expression for the controller gain as a function of the graph Laplacian eigenvalues has also been derived. The practical applications of this work have been demonstrated through simulations</div>

<div>This thesis presents work and simulations containing the use of Artificial Intelligence for Unmanned Aerial Vehicles in search and rescue and/or surveillance operations. The goal is to create a vision system that leverages Artificial Intelligence, mainly Deep Learning techniques to build a pipeline that enables fast and accurate classification of the environment of the robot. Deep Neural Networks are trained and tested on ’emergency situational data. Further, the power of this vision system is leveraged to extend the problem onto a multiagent system to handle fault tolerance. The multi-agent system is also made resilient to Byzantine malicious attacks to help improve the reliability of the system.</div><div><br></div><div>This thesis also shows the use of Artificial Intelligence for effective surveillance for defense related purposes. Tracking the GPS coordinates of a boat using only the video of the boat captured by a camera and the GPS coordinates of the camera itself is demonstrated. The solution was tested by the Department of Defense - Department of the Navy, Naval Information Warfare Center Pacific.</div>

<p> </p> <p>This work proposes a proof of concept along with a prototype of a multi-agent autonomous drone system that can be used to detect, and capture a intruding adversarial drone. The functional Counter Unmanned Aerial System (CUAS) prototype is used to convey the feasibility of a persistent multi-agent aerial protection and apprehension system by demonstrating important features of the mission through both simulation and field testing.<br> </p> <p> </p>

<div>Class 8 trucks on average drive the most miles and consume the most fuel of any major vehicle category annually. Indiana specifically is the fifth busiest state for commercial freight traffic and moves $750 billion dollars of freight annually, and this number is expected to grow by 60% by 2040. Reducing fuel consumption for class 8 trucks would have a significant benefit on business and the proportional decrease in CO₂ would be exceptionally beneficial for the environment.</div><div><br></div><div>Platooning is one of the most important strategies for increasing class 8 truck fuel savings. Platooning alone can help trucks save upwards of 7% platoon average fuel savings on flat ground. However, it can be difficult for a platooning controller to maintain a desired truck separation during uncoordinated shifting events. Using a high-fidelity simulation model, it is shown that simultaneous shifting–having the follow truck shift whenever the lead truck shifts (unless shifting would cause its engine to overspeed or underspeed)–decreases maximum truck separation by 24% on a moderately challenging grade route and 40% on a heavy grade route.</div><div><br></div><div>Model Predictive Control (MPC) of the follow truck is considered as a means to reduce the distance the follow truck falls behind during uncoordinated shifting events. The result in simulation is a reduction in maximum truck separation of 1% on a moderately challenging grade route and 19% on a heavy grade route. However, simultaneous shifting largely alleviates the need for MPC for the sake of tracking for the follow truck.</div><div><br></div><div>A different MPC formulation is considered to dynamically change the desired set point for truck separation for routes through a strategy called Route Optimized Gap Growth (ROGG). The result in simulation is 1% greater fuel savings on a moderately challenging grade route and 7% greater fuel savings on a route with heavy grade for the follow truck.</div>

<div>Ensuring the robustness and resilience of safety-critical systems from civil aviation to military surveillance technologies requires improvements to target tracking capabilities. Implementing target tracking as a distributed function can improve the quality and availability of information for end users. Any errors in the model of a target's dynamics or a sensor network's measurement process will result in estimates with degraded accuracy or even filter divergence. This dissertation solves a distributed estimation problem for estimating the state of a dynamical system and the parameters defining a model of that system.</div><div>The novelty of this work lies in the ability of a sensor network to maintain consensus on state and parameter estimates through local communications between sensor platforms.</div>

<div>Efficient engine operation is a fundamental control problem in automotive applications. Robust control algorithms are necessary to achieve satisfactory, safe engine performance</div><div>at all operating conditions while reducing emissions. This thesis develops a framework for control architecture design to enable robust air handling system management.</div><div><br></div><div>The first work in the thesis derives a simple physics-based, control-oriented model for turbocharged lean burn engines which is able to capture the critical engine dynamics that are</div><div>needed to design the controller. The control-oriented model is amenable for control algorithm development and includes the impacts of modulation to any combination of four actuators: throttle valve, bypass valve, fuel rate, and wastegate valve. The controlled outputs: engine speed, differential pressure across throttle and air-to-fuel ratio are modeled as functions of selected states and inputs. Two validation strategies, open-loop and closed-loop are used to validate the accuracy of both nonlinear and linear versions of the control-oriented model. The relative gain array is applied to the linearized engine model to understand the degree of interactions between plant inputs and outputs as well as the best input-output pairing as a function of frequency. With strong evidence of high degree of coupling between inputs and outputs, a coordinated multiple-input multiple-output (MIMO) controller is hypothesized to perform better than a single-input single-output (SISO) controller. A framework to design robust model-based H1 MIMO controllers for any given linear plant, while considering state and output multiplicative uncertainties as well as actuator bandwidths is developed. The framework also computes the singular structure value, μ for the uncertain closed-loop system to quantify robustness, both in terms of stability and performance. The multi-tracking control problem targets engine speed, differential pressure across throttle as well as air-to-fuel ratio to achieve satisfactory engine performance while also preventing compressor surge and reducing engine emissions. A controller switching methodology using slow-fast controller decomposition and hysteresis at switching points is proposed to smoothly switch control authority between several MIMO controllers. The control design approach is applied to a truth-reference GT-Power engine model to evaluate the closed-loop controller performance. The engine response obtained using the robust MIMO controller is compared with that obtained using a state-of-the-art benchmark controller to evaluate the additional benefits of the MIMO controller.</div><div><br></div><div><div>In the second study, a robust 2-degree of freedom controller that commands eBooster speed to control air-to-fuel ratio, and a robust MIMO coordinated controller to control gas</div><div>exchange process in a diesel engine with electrified air handling architecture are developed. The MIMO controller simultaneously controls engine speed, mass fraction of the recirculated exhaust gas as well as air-to-fuel ratio. The actuators available for control in the engine include the exhaust gas recirculation valve, exhaust throttle valve, fuel injection rate, eBooster speed, eBooster bypass valve. To design the robust eBooster controller, the input-output relationship between eBooster speed and air-to-fuel ratio is estimated using system identification techniques. The robust MIMO controller is synthesized using a physics-based mean value control-oriented engine model that accurately represents the high-fidelity GT-Power model. In the first control strategy, the robust eBooster controller is added to an already existing stock engine control unit while in the second control strategy, the stock engine control unit is replaced with the multiple-input multiple-output controller. The two control architectures are tested under different operating conditions to evaluate the controller performance. Simulation results with the control architectures developed in the thesis are compared to a baseline engine configuration, where the engine operates without eBooster. Although it is observed that both these control algorithms significantly improve engine performance as compared to the baseline configuration, MIMO controller provides the best engine performance overall.</div></div>