UMD Theses and Dissertations
Permanent URI for this collectionhttp://hdl.handle.net/1903/3
New submissions to the thesis/dissertation collections are added automatically as they are received from the Graduate School. Currently, the Graduate School deposits all theses and dissertations from a given semester after the official graduation date. This means that there may be up to a 4 month delay in the appearance of a given thesis/dissertation in DRUM.
More information is available at Theses and Dissertations at University of Maryland Libraries.
Browse
2 results
Search Results
Item Getting on the Same Page: How Leaders Build Trust Consensus in Teams and Its Consequences(2012) Fulmer, C. Ashley; Ostroff, Cheri; Psychology; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Existing organizational research has demonstrated that team members' trust in leaders is positively related to a team's bottom-line outcomes. However, little is known about how collective trust in leaders develops among team members. To address this gap, the present study examines the effects of multiple emergent processes on the extent to which team members exhibit consensus in trust in their leader. In particular, it was proposed that the most important factors for the emergence, and the degree of consensus, of collective trust in leaders should have the same referent target as the collective construct (i.e., the leader) and concern behaviors that involve interactions between the leader and team members. Thus, the leader behavior and interactions variables of showing concern, leading by example, and monitoring were expected to exert stronger influence on the consensus in trust in leaders than leader attributes (ability and integrity) and team factors (open communication and demographic diversity). Further, the degree of consensus in trust in leaders was predicted to have both an independent and interaction effect with the mean level of trust in leaders in influencing team performance and voice behavior. Three waves of survey data were collected from teams with new leadership in a large academic military institution. Data from 719 team members from 105 teams were used to test these predictions by analyzing consensus concurrently and changes in consensus over time. The results generally supported the relative importance of leader showing concern and leading by example on the degree of consensus in trust in leaders in the concurrent model. For changes in consensus, leading by example was particularly important. In addition, while consensus was not independently related to the team performance and voice behaviors, it interacted with the mean level in influencing the outcomes in both the concurrent and change models. Taken together, the findings suggest that some leader behaviors are important for the development of collective trust or consensus in trust in leaders, and further suggest that consensus can act as a boundary condition for the effect of the mean level of trust in leaders on team outcomes. By focusing on the consensus in trust in leaders, this research begins to shed light on how consensus in trust develops among team members with respect to their leader and has implications for understanding trust, leadership, and emergence.Item Consensus problems and the effects of graph topology in collaborative control(2009) Hovareshti, Pedram; Baras, John S; Electrical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)In this dissertation, several aspects of design for networked systems are addressed. The main focus is on combining approaches from system theory and graph theory to characterize graph topologies that result in efficient decision making and control. In this framework, modelling and design of sparse graphs that are robust to failures and provide high connectivity are considered. A decentralized approach to path generation in a collaborative system is modelled using potential functions. Taking inspiration from natural swarms, various behaviors of the system such as target following, moving in cohesion and obstacle avoidance are addressed by appropriate encoding of the corresponding costs in the potential function and using gradient descent for minimizing the energy function. Different emergent behaviors emerge as a result of varying the weights attributed with different components of the potential function. Consensus problems are addressed as a unifying theme in many collaborative control problems and their robustness and convergence properties are studied. Implications of the continuous convergence property of consensus problems on their reachability and robustness are studied. The effects of link and agent faults on consensus problems are also investigated. In particular the concept of invariant nodes has been introduced to model the effect of nodes with different behaviors from regular nodes. A fundamental association is established between the structural properties of a graph and the performance of consensus algorithms running on them. This leads to development of a rigorous evaluation of the topology effects and determination of efficient graph topologies. It is well known that graphs with large diameter are not efficient as far as the speed of convergence of distributed algorithms is concerned. A challenging problem is to determine a minimum number of long range links (shortcuts), which guarantees a level of enhanced performance. This problem is investigated here in a stochastic framework. Specifically, the small world model of Watts and Strogatz is studied and it is shown that adding a few long range edges to certain graph topologies can significantly increase both the rate of convergence for consensus algorithms and the number of spanning trees in the graph. The simulations are supported by analytical stochastic methods inspired from perturbations of Markov chains. This approach is further extended to a probabilistic framework for understanding and quantifying the small world effect on consensus convergence rates: Time varying topologies, in which each agent nominally communicates according to a predefined topology, and switching with non-neighboring agents occur with small probability is studied. A probabilistic framework is provided along with fundamental bounds on the convergence speed of consensus problems with probabilistic switching. The results are also extended to the design of robust topologies for distributed algorithms. The design of a semi-distributed two-level hierarchical network is also studied, leading to improvement in the performance of distributed algorithms. The scheme is based on the concept of social degree and local leader selection and the use of consensus-type algorithms for locally determining topology information. Future suggestions include adjusting our algorithm towards a fully distributed implementation. Another important aspect of performance in collaborative systems is for the agents to send and receive information in a manner that minimizes process costs, such as estimation error and the cost of control. An instance of this problem is addressed by considering a collaborative sensor scheduling problem. It is shown that in finding the optimal joint estimates, the general tree-search solution can be efficiently solved by devising a method that utilizes the limited processing capabilities of agents to significantly decrease the number of search hypotheses.