CAREER: Agent-Based Decentralized Control of Distributed Systems

Vanderbilt University, Nashville TN

Investigators

Abstract

The purpose of the proposed effort is to investigate the control of distributed systems (i.e. systems with distributed mass, stiffness or other dynamic properties) in the presence of decentralized, networked, embedded information systems that deeply penetrate the system to be controlled. The networked imbedded system consists of numerous nodes, each consisting of a small microprocessor, sensor, and actuator. Furthermore, all of the nodes will be capable of communicating over a wired or wireless network. This configuration means that each node is subject to the following constraints: 1) limited computational power and memory; 2) limited information as to overall system response; and 3) limited authority over the physical system. The primary challenge of this program will be the development of decentralized control algorithms capable of performing successfully under such conditions. The research objectives will be accomplished through a parallel theoretical/experimental effort. Theoretical work will focus on the development of simulation tools and decentralized control algorithms that meet the outlined objectives. The experimental efforts will seek to demonstrate this technology on a series of increasingly complex systems. The research program promises to develop new analysis tools and yield new insight into the behavior of decentralized control systems and, in particular, their application to distributed parameter systems. Among the many new contributions to the field will be: the investigation and development of decentralized control of distributed systems; improved knowledge on the global behavior of decentralized systems defined by localized actions; methods for constructing global performance metrics based on decentralized sensor processing; and the development of adaptive control algorithms for the creation of desirable global performance based on localized information. The primary goal of the educational program is to lead the development of a Mechanical Engineering curriculum that synergistically integrates a variety of learning/teaching approaches in order to reach a more diverse student population. While traditional lecture formats address the majority of individual learning modes, the introduction of non-traditional approaches will not only enhance current student's learning but also attract and enhance the retention of a more diverse student population who's learning needs are not currently met. The goal will be met by introducing several new elements to those aspects of the curriculum under the PIs' purview. These include alternative lecture teaching techniques, research oriented laboratory exercises, integration of graduate research methods and equipment into undergraduate teaching labs, and improved research opportunities for underrepresented individuals.