Graduation Year

2019

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

Eduardo L. Pasiliao, Ph.D.

Committee Member

Rajiv Dubey, Ph.D.

Committee Member

Kyle Reed, Ph.D.

Committee Member

Yasin Yilmaz, Ph.D.

Keywords

Finite-time Control, Formation Control, Multiplex Networks, Nullspace Control, Sensor Networks

Abstract

The first objective of this dissertation is to develop novel distributed control architectures allowing spatiotemporal control of multiagent systems as applied to formation control. In addition, its second objective is to introduce distributed estimation frameworks for dynamic information fusion for addressing the heterogeneity in sensor networks.

Changing the spatial and temporal properties of agent teams in a distributed manner and in real-time is an open problem in the control system literature as multiagent systems are often required to complete tasks with ever-increasing complexity in adverse conditions and dynamic environments. Motivated by this standpoint, this dissertation aims to address challenges related to spatiotemporal control of multiagent systems by proposing three novel tools and methods: The multiplex information networks; the nullspace control; and the time transformation method. First, existing distributed control algorithms utilize only a single layer information exchange rule leading to the multiagent systems have fixed spatial and temporal properties (e.g., the size, orientation and spatial evolution rate of a formation are fixed). To this end, we introduce multiplex information networks with multiple information exchange layers comprising both intralayer and interlayer communication links to allow the spatial and temporal properties of multiagent systems (e.g., the formation’s size, orientation, and bandwidth) being manipulated in a distributed manner. Moreover, tools and methods from differential potential fields are used for connectivity maintenance and collision avoidance between agents. Second, complex cooperative behaviors in multiagent systems are

restricted under existing control architectures since the local interactions between agents are encoded in the standard Laplacian matrix, which has the nullspace spanning the vector of ones. A novel method proposed in this dissertation defines a more general version for Laplacian matrix, whose nullspace can be manipulated as desired, and reveals a better understanding of the local interactions as well as allows a broader range of complex cooperative behaviors in multiagent systems. Third, time-critical applications, where a task is required to be completed at a user-defined convergence time, is another challenge. Distributed control algorithms for such applications are also developed and generalized. Specifically, a novel time transformation method is employed to transform the system from the prescribed time interval t in [0;T) to an equivalent system over the stretched infinite-time interval s in [0;\infty) for analysis purposes.

One additional challenge in multiagent system is the heterogeneity in sensor networks, which prevents dynamics information correctly being fused. Therefore, another major contribution of this dissertation is to introduce and analyze new distributed input and state estimation architectures for addressing the heterogeneity in sensor networks stemming from different in sensor modalities, quality of sensing information (value of information), and information roles of nodes (active and passive). Both fixed and time-varying information roles of nodes are investigated. Furthermore, some existing literature (see, for example, [2]) implicitly considers nodes with different information roles, yet they require global sufficient stability conditions for analysis while our proposed architectures only utilize local measurements and information both in execution and design stages to guarantee the stability and performance of the overall sensor network.

Finally, the stability of the proposed architectures is theoretically analyzed and their efficacy is illustrated on numerical examples as well as verified with experiments on various mobile robot platforms.

Download

Share

COinS