Dynamic lines of collaboration in e-Work systems
Abstract
E-Work expands in the rapid evolution of intelligent technologies for interaction, communication, and collaboration. Enabled by e-Work, networked agents can provide services collaboratively to interdependent clients, i.e., the Network-to-Network (N2N) services. The complex dependencies in networks challenge agents to maintain efficiency in their distributed yet collaborative operations. To gain systematic understanding of N2N services in e-Work, and to improve the operations of N2N services, the Dynamic Lines of Collaboration (DLOC) model is proposed, developed, explored and validated in this dissertation. DLOC model captures the dynamics of the changing structure of a service team while performing inter-related services on and throughout a network of clients. The theoretical foundation of DLOC lies in complex network theory, scheduling theory, and collaborative control theory. Based on theoretical analysis of DLOC, three protocols are designed to effectively control the collaboration of agents: (1) Asynchronous Collaboration Requirement Planning (ACRP) is used to design the configuration of a team of agents prepared to handle collaborative tasks. (2) Centrality-Based depot Allocation (CBA) protocol efficiently aligns the service team with the client network to provide better coverage of services. (3) Neuroplasticity-inspired scheduling protocols determine the response operations to minimize the total latency of all tasks. Three classes of critical performance metrics, i.e., quality, time and cost, are designed to evaluate the N2N services controlled by the protocols. To validate the DLOC model and protocols, a new version of Teamwork Integration Evaluator (TIE) software - TIE/DLOC is developed. The evaluator is able to analyze the structures of a client network and a service team, then apply different control protocols to the e-Work system, and finally evaluate several protocols based on their performance in N2N services. In this dissertation, the applications of the DLOC model and control protocols are discussed, experimented, and evaluated. The first set of experiments is on the configuration design of Reconfigurable End-Effectors (REEs) for automated harvesting systems. Different configurations of REE components provide various grasp qualities on diverse targets. ACRP is used to determine the configuration network of REE so that the quality and yield of harvesting operations for various vegetables are optimized. The second set of experiments is on the collaborative response operations during disruptions in the critical Cyber-Physical Infrastructure (CPI). When disruptions occur in a CPI, cascading failures may lead to catastrophic damage. The disruption response operations are crucial, because if the responders can provide the correct service to the right components of CPI at the right time, failures and damages can be limited and kept within a relatively small range. CBA and neuroplasticity-inspired scheduling protocols developed in this research facilitate the allocation of responders and online task scheduling to minimize the latency before CPI components are repaired. This experiment is based on both simulated CPI networks and real networks of water supply and power grid. Other potential applications of the DLOC model and control protocols are discussed, including: collaborative visual analytics, heterogeneous robotic material handling, and quality assurance planning in manufacturing. By applying the DLOC model and TIE/DLOC software on the aforementioned e-Work systems for experimental studies, the performance under different control protocols are obtained. The experimental results show that there is a significant performance improvement by the control protocols developed in this research compared with other conventional design, allocation, and scheduling protocols in practice. The improvement implies that through network awareness, i.e., by exploiting the structures of both the client and the server networks for the control of N2N services a higher level of collaborative intelligence can be achieved. Based on this dissertation, future research can advance the effectiveness of emerging e-Work systems with additional protocols and decision dimensions.
Degree
Ph.D.
Advisors
Nof, Purdue University.
Subject Area
Industrial engineering|Systems science|Operations research
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.