The efficiency, security and resiliency are very important factors for the operation of a distribution power system. Taking into account customer demand and energy resource constraints, electric utilities not only need to provide reliable services but also need to operate a power grid as efficiently as possible. The objective of this dissertation is to design, develop and deploy the Multi-Agent Systems (MAS) - together with control algorithms - that enable demand response (DR) implementation at the customer level, focusing on both residential and commercial customers. For residential applications, the main objective is to propose an approach for a smart distribution transformer management. The DR objective at a distribution transformer is to ensure that the instantaneous power demand at a distribution transformer is kept below a certain demand limit while impacts of demand restrike are minimized. The DR objectives at residential homes are to secure critical loads, mitigate occupant comfort violation, and minimize appliance run-time after a DR event. For commercial applications, the goal is to propose a MAS architecture and platform that help facilitate the implementation of a Critical Peak Pricing (CPP) program. Main objectives of the proposed DR algorithm are to minimize power demand and energy consumption during a period that a CPP event is called out, to minimize occupant comfort violation, to minimize impacts of demand restrike after a CPP event, as well as to control the device operation to avoid restrikes. Overall, this study provides an insight into the design and implementation of MAS, together with associated control algorithms for DR implementation in smart buildings. The proposed approaches can serve as alternative solutions to the current practices of electric utilities to engage end-use customers to participate in DR programs where occupancy level, tenant comfort condition and preference, as well as controllable devices and sensors are taken into account in both simulated and real-world environments. Research findings show that the proposed DR algorithms can perform effectively and efficiently during a DR event in residential homes and during the CPP event in commercial buildings. / Ph. D.

This work addresses applications of search theory where a mobile search agent seeks to find an unknown number of stationary targets randomly distributed in a bounded search domain. We assume that the search mission is subject to a time or distance constraint, and that the local environmental conditions affect sensor performance. Because the environment varies by location, the effectiveness of the search sensor also varies by location. Our contribution to search theory includes new decision-theoretic approaches for generating optimal search plans in the presence of false alarms and uncertain environmental variability. We also formally define the value of environmental information for improving the effectiveness of a search mission, and we develop methods for optimal deployment of assets that can acquire environmental information in order to improve search effectiveness. Finally, we extend our research to the case of multiple cooperating search agents. For the case that inter-agent communication is severely bandwidth-limited, such as in subsea applications, we propose a method for assessing the expected value of inter-agent communication relative to joint search effectiveness. Our results lead to a method for determining when search agents should communicate. Our contributions to search theory address important applications that range from subsea mine-hunting to post-disaster search and rescue applications. / PHD / We address search applications where a mobile search agent seeks to find an unknown number of stationary targets randomly distributed in a bounded search domain. The search agent is equipped with a search sensor that detects the targets at a location. Sensor measurements are often imperfect due to possible missed detections and false alarms. We also consider that the local environmental conditions affect the quality of the data acquired from the search sensor. For instance, if we are searching for a target that has a rocky shape, we expect that it will be harder to find that target in a rocky environment. We consider that the search mission is subject to a time or distance constraint, and thus, search can be performed on only a subset of locations. Our goal in this study is to formally determine where to acquire the search measurements so that the search effectiveness can be maximized. We also formally define the value of acquiring environmental information for improving the effectiveness of a search mission, and we develop methods for optimal deployment of assets that can acquire environmental information in order to improve search effectiveness. Finally, we address the cases where multiple search assets collaboratively search the environment and they can communicate their local information with each other. We are particularly interested in determining when a vehicle should communicate with another vehicle so that the joint search effectiveness can be improved. Our contributions to search theory address important applications that range from subsea mine-hunting to post-disaster search and rescue applications.

This dissertation addresses a fundamental question of multi-agent coordination: what infor- mation should be sent to whom and when, with the limited resources available to each agent? Communication requirements for multi-agent systems can be rather high when an accurate picture of the environment and the state of other agents must be maintained. To reduce the impact of multi-agent coordination on networked systems, e.g., power and bandwidth, this dissertation introduces new concepts to enable Intelligent Knowledge Distribution (IKD), including Constrained-action POMDPs (CA-POMDP) and concurrent decentralized (CoDec) POMDPs for an agnostic plug-and-play capability for fully autonomous systems. Each agent runs a CoDec POMDP where all the decision making (motion planning, task allocation, asset monitoring, and communication) are separated into concurrent individual MDPs to reduce the combinatorial explosion of the action and state space while maintaining dependencies between the models. We also introduce the CA-POMDP with action-based constraints on partially observable Markov decision processes, rewards driven by the value of information, and probabilistic constraint satisfaction through discrete optimization and Markov chain Monte Carlo analysis. IKD is adapted real-time through machine learning of the actual environmental impacts on the behavior of the system, including collaboration strategies between autonomous agents, the true value of information between heterogeneous systems, observation probabilities and resource utilization. / Doctor of Philosophy / This dissertation addresses a fundamental question behind when multiple autonomous sys- tems, like drone swarms, in the field need to coordinate and share data: what information should be sent to whom and when, with the limited resources available to each agent? Intelligent Knowledge Distribution is a framework that answers these questions. Communication requirements for multi-agent systems can be rather high when an accurate picture of the environment and the state of other agents must be maintained. To reduce the impact of multi-agent coordination on networked systems, e.g., power and bandwidth, this dissertation introduces new concepts to enable Intelligent Knowledge Distribution (IKD), including Constrained-action POMDPs and concurrent decentralized (CoDec) POMDPs for an agnostic plug-and-play capability for fully autonomous systems. The IKD model was able to demonstrate its validity as a "plug-and-play" library that manages communications between agents that ensures the right information is being transmitted at the right time to the right agent to ensure mission success.

Optimizing connected automated vehicle movements through roadway intersections is a challenging problem. Traditional traffic control strategies, such as traffic signals are not optimal, especially for heavy traffic. Alternatively, centralized automated vehicle control strategies are costly and not scalable given that the ability of a central controller to track and schedule the movement of hundreds of vehicles in real-time is highly questionable. In this research, a series of fully distributed heuristic algorithms are proposed where vehicles in the vicinity of an intersection continuously cooperate with each other to develop a schedule that allows them to safely proceed through the intersection while incurring minimum delays. An algorithm is proposed for the case of an isolated intersection then a number of algorithms are proposed for a network of intersections where neighboring intersections communicate directly or indirectly to help the distributed control at each intersection makes a better estimation of traffic in the whole network. An algorithm based on the Godunov scheme outperformed optimized signalized control. The simulated experiments show significant reductions in the average delay. The base algorithm is successfully added to the INTEGRATION micro-simulation model and the results demonstrate improvements in delay, fuel consumption, and emissions when compared to roundabout, signalized, and stop sign controlled intersections. The study also shows the capability of the proposed technique to favor emergency vehicles, producing significant increases in mobility with minimum delays to the other vehicles in the network. / Ph. D.

Unfavorable environmental and (or) human displacement may engender the need for Search and Rescue (SAR). Challenges such as inaccessibility, large search areas, and heavy reliance on available responder count, limited equipment and training makes SAR a challenging problem. Additionally, SAR operations also pose significant risk to involved responders. This opens a remarkable opportunity for robotic systems to assist and augment human understanding of the harsh environments. A large body of work exists on the introduction of ground and aerial robots in visual and temporal inspection of search areas with varying levels of autonomy. Unfortunately, limited autonomy is the norm in such systems, due to the limitations presented by on-board UAV resources and networking capabilities. In this work we propose a new multi-agent approach to SAR and introduce a wearable compute cluster in the form factor of a backpack. The backpack allows offloading compute intensive tasks such as Lost Person Behavior Modelling, Path Planning and Deep Neural Network based computer vision applications away from the UAVs and offers significantly high performance computers to execute them. The backpack also provides for a strong networking backbone and task orchestrators which allow for enhanced coordination and resource sharing among all the agents in the system. On the basis of our benchmarking experiments, we observe that the backpack can significantly boost capabilities and success in modern SAR responses. / Master of Science / Unfavorable environmental and (or) human displacement may engender the need for Search and Rescue (SAR). Challenges such as inaccessibility, large search areas, and heavy reliance on available responder count, limited equipment and training makes SAR a challenging problem. Additionally, SAR operations also pose significant risk to involved responders. This opens a remarkable opportunity for robotic systems to assist and augment human understanding of the harsh environments. A large body of work exists on the introduction of ground and aerial robots in visual and temporal inspection of search areas with varying levels of autonomy. Unfortunately, limited autonomy is the norm in such systems, due to the limitations presented by on-board UAV resources and networking capabilities. In this work we propose a new multi-agent approach to SAR and introduce a wearable compute cluster in the form factor of a backpack. The backpack allows offloading compute intensive tasks such as Lost Person Behavior Modelling, Path Planning and Deep Neural Network based computer vision applications away from the UAVs and offers significantly high performance computers to execute them. The backpack also provides for a strong networking backbone and task orchestrators which allow for enhanced coordination and resource sharing among all the agents in the system. On the basis of our benchmarking experiments, we observe that the backpack can significantly boost capabilities and success in modern SAR responses.

Generalized collaborative drones are a technology that has many potential benefits. General purpose drones that can handle exploration, navigation, manipulation, and more without having to be reprogrammed would be an immense breakthrough for usability and adoption of the technology. The ability to develop these multi-task, multi-agent drone systems is limited by the lack of available training environments, as well as deficiencies of multi-task learning due to a phenomenon known as catastrophic forgetting. In this thesis, we present a set of simulation environments for exploring the abilities of multi-task drone systems and provide a platform for testing agents in incremental single-agent and multi-agent learning scenarios. The multi-task platform is an extension of an existing drone simulation environment written in Python using the PyBullet Physics Simulation Engine, with these environments incorporated. Using this platform, we present an analysis of Incremental Learning and detail the beneficial impacts of using the technique for multi-task learning, with respect to multi-task learning speed and catastrophic forgetting. Finally, we introduce a novel algorithm, Incremental Learning with Second-Order Approximation Regularization (IL-SOAR), to mitigate some of the effects of catastrophic forgetting in multi-task learning. We show the impact of this method and contrast the performance relative to a multi-agent multi-task approach using a centralized policy sharing algorithm. / Master of Science / Machine Learning techniques allow drones to be trained to achieve tasks which are otherwise time-consuming or difficult. The goal of this thesis is to facilitate the work of creating these complex drone machine learning systems by exploring Reinforcement Learning (RL), a field of machine learning which involves learning the correct actions to take through experience. Currently, RL methods are effective in the design of drones which are able to solve one particular task. The next step in this technology is to develop RL systems which are able to handle generalization and perform well across multiple tasks. In this thesis, simulation environments for drones to learn complex tasks are created, and algorithms which are able to train drones in multiple hard tasks are developed and tested. We explore the benefits of using a specific multi-task training technique known as Incremental Learning. Additionally, we consider one of the prohibitive factors of multi-task machine learning-based solutions, the degradation problem of agent performance on previously learned tasks, known as catastrophic forgetting. We create an algorithm that aims to prevent the impact of forgetting when training drones sequentially on new tasks. We contrast this approach with a multi-agent solution, where multiple drones learn simultaneously across the tasks.

The security and resiliency of electric power supply to serve critical facilities are of high importance in today's world. Instead of building large electric power grids and high capacity transmission lines, an intelligent microgrid (or smart grid) can be considered as a promising power supply alternative. In recent years, multi-agent systems have been proposed to provide intelligent energy control and management systems in microgrids. Multi-agent systems offer their inherent benefits of flexibility, extensibility, autonomy, reduced maintenance and more. The implementation of a control network based on multi-agent systems that is capable of making intelligent decisions on behalf of the user has become an area of intense research. Many previous works have proposed multi-agent system architectures that deal with buying and selling of energy within a microgrid and algorithms for auction systems. The others proposed frameworks for multi-agent systems that could be further developed for real life control of microgrid systems. However, most proposed methods ignore the process of sharing energy resources among multiple distinct sets of prioritized loads. It is important to study a scenario that emphasizes on supporting critical loads during outages based on the user's preferences and limited capacity. The situation becomes further appealing when an excess DER capacity after supplying critical loads is allocated to support non-critical loads that belong to multiple users. The previous works also ignore the study of dynamic interactions between the agents and the physical systems. It is important to study the interaction and time delay when an agent issues a control signal to control a physical device in a microgrid and when the command is executed. Agents must be able to respond to the information sensed from the external environment quickly enough to manage the microgrid in a timely fashion. The ability of agents to disconnect the microgrid during emergencies should also be studied. These issues are identified as knowledge gaps that are of focus in this thesis. The objective of this research is to design, develop and implement a multi-agent system that enables real-time management of a microgrid. These include securing critical loads and supporting non-critical loads belonging to various owners with the distributed energy resource that has limited capacity during outages. The system under study consists of physical (microgrid) and cyber elements (multi-agent system). The cyber part or the multi-agent system is of primary focus of this work. The microgrid simulation has been implemented in Matlab/Simulink. It is a simplified distribution circuit that consists of one distributed energy resources (DER), loads and the main grid power supply. For the multi-agent system implementation, various open source agent building toolkits are compared to identify the most suitable agent toolkit for implementation in the proposed multi-agent system. The agent architecture is then designed by dividing overall goal of the system into several smaller tasks and assigning them to each agent. The implementation of multi-agent system was completed by identifying Roles (Role Modeling) and Responsibilities (Social and Domain Responsibilities) of agents in the system, and modeling the Knowledge (Facts), rules and ontology for the agents. Finally, both microgrid simulation and multi-agent system are connected together via TCP/IP using external java programming and a third party TCP server in the Matlab/Simulink environment. In summary, the multi-agent system is designed, developed and implemented in several simulation test cases. It is expected that this work will provide an insight into the design and development of a multi-agent system, as well as serving as a basis for practical implementation of an agent-based technology in a microgrid environment. Furthermore, the work also contributes to new design schemes to increase multi-agent system's intelligence. In particular, these include control algorithms for intelligently managing the limited supply from a DER during emergencies to secure critical loads, and at the same time supporting non-critical loads when the users need the most. / Master of Science

One of the major planning problems faced by medium and large manufacturing enterprises is the distribution of production over various (production) facilities. The need for cross-facility capacity management is most evident in the high-tech industries having capital-intensive equipment and short technology life cycle. There have been solutions proposed in the literature that are based on the lagragian decomposition method which separate the overall multiple product problem into a number of single product problems. We believe that multi-agent systems, given their distributed problem solving approach can be used to solve this problem, in its entirety, more effectively. According to other researchers who have worked in this field, auction theoretic mechanisms are a good way to solve complex production planning problems. This research study develops a multi-agent system and negotiation protocol based on combinatorial auction framework to solve the given multi-facility planning problem. The output of this research is a software library, which can be used as a multi-agent system model of the multi-product, multi-facility capacity allocation problem. The negotiation protocol for the agents is based on an iterative combinatorial auction framework which can be used for making allocation decisions in this environment in real-time. A simulator based on this library is created to validate the multi-agent model as well as the auction theoretic framework for different scenarios in the problem domain. The planning software library is created using open source standards so that it can be seamlessly integrated with scheduling library being developed as a part of the Advanced Planning and Scheduling (APS) system project or any other software suite which might require this functionality. The research contribution of this study is in terms of a new multi-agent architecture for an Advanced Planning and Control (APS) system as well as a novel iterative combinatorial auction mechanism which can be used as an agent negotiation protocol within this architecture. The theoretical concepts introduced by this research are implemented in the MultiPlanner production planning tool which can be used for generating master production plans for manufacturing enterprises. The validation process carried out on both the iterative combinatorial framework and the agent-based production planning methodology demonstrate that the proposed solution strategies can be used for integrated decision making in the multi-product, multi-facility production planning domain. Also, the software tool developed as part of this research is a robust, platform independent tool which can be used by manufacturing enterprises to make relevant production planning decisions. / Master of Science

Pursuit-evasion problems comprise a set of pursuers that strive to catch oneor several evaders, often in a constrained environment. This thesis proposesand compares heuristic algorithms for pursuit-evasion problems wherein several double integrator agents pursue a single evader in a bounded subset of theEuclidean plane. Different methods for assigning surrounding target points tothe pursuers are tested numerically. In addition, a method which finds the timeoptimal strategy for pursuing a static target in an unconstrained setting is presented, and is then used to pursue the assigned, dynamic, target. Numericalresults show that the time optimal strategy for pursuing a static target translateswell to the dynamic problem.

This thesis addresses the problem of positional consensus of multi-agent systems. A positional consensus is achieved when the agents converge to a point. Some applications of this class of problem is in mid-air refueling of the aircraft or UAVs, targeting a geographical location, etc. In this research work some positional consensus algorithms have been developed. They can be categorized in two part (i) Broadcast control based algorithm (ii) Distributed control based algorithm. In case of broadcast based algorithm control strategies for a group of agents is developed to achieve positional consensus. The problem is constrained by the requirement that every agent must be given the same control input through a broadcast communication mechanism. Although the control command is computed using state information in a global framework, the control input is implemented by the agents in a local coordinate frame. The mathematical formulation has been done in a linear programming framework that is computationally less intensive than earlier proposed methods. Moreover, a random perturbation input in the control command, that helps to achieve reasonable proximity among agents even for a large number of agents, which was not possible with the existing strategy in the literature, is introduced. This method is extended to achieve positional consensus at a pre-specified location. A comparison between the LP approach and the existing SOCP based approach is also presented. Some of the algorithm has been demonstrated successfully on a robotic platform made from LEGO Mindstorms NXT Robots. In the second case of broadcast based algorithm, a decentralized algorithm for a group of multiple autonomous agents to achieve positional consensus has been developed using the broadcast concept. Even here, the mathematical formulation has done using a linear programming framework. Each agent has some sensing radius and it is capable of sensing position and orientation with other agents within their sensing region. The method is computationally feasible and easy to implement. In case of distributed algorithms, a computationally efficient distributed rendezvous algorithm for a group of autonomous agents has been developed. The algorithm uses a rectilinear decision domain (RDD), as against the circular decision domain assumed in earlier work available in the literature. This helps in reducing its computational complexity considerably. An extensive mathematical analysis has been carried out to prove the convergence of the algorithm. The algorithm has also been demonstrated successfully on a robotic platform made from LEGO Mindstorms NXT Robots.