http://hdl.handle.net/1721.1/7905 Sun, 26 May 2013 06:26:26 GMT 2013-05-26T06:26:26Z http://hdl.handle.net/1721.1/78148 Approximating the performance of a last mile transportation system Wang, Hai, S.M. Massachusetts Institute of Technology The Last Mile Problem (LMP) refers to the provision of travel service from the nearest public transportation node to a home or office. We study the supply side of this problem in a stochastic setting, with batch demands resulting from the arrival of groups of passengers at rail stations or bus stops who request last-mile service. Closed-form bounds and approximations are derived for the performance of Last Mile Transportations Systems as a function of the fundamental design parameters of such systems. An initial set of results is obtained for the case in which a fleet of vehicles of unit-capacity provides the Last Mile service and each delivery route consists of a simple round-trip between the rail station and bus stop and the single passenger's destination. These results are then extended to the general case in which the capacity of a vehicle is an arbitrary, but typically small (under 10) number. It is shown through comparisons with simulation results, that a particular strict upper bound and an approximate upper bound, both derived under similar assumptions, perform consistently and remarkably well for the entire spectrum of input values and conditions simulated. These expressions can therefore be used for the preliminary planning and design of Last Mile Transportation Systems, especially for determining approximately resource requirements, such as the number of vehicles/servers needed to achieve some pre-specified level of service. Thesis (S.M. in Transportation)--Massachusetts Institute of Technology, Dept. of Civil and Environmental Engineering; and, (S.M.)--Massachusetts Institute of Technology, Sloan School of Management, Operations Research Center, 2012.; Cataloged from PDF version of thesis.; Includes bibliographical references (p. 113). Sun, 01 Jan 2012 00:00:00 GMT http://hdl.handle.net/1721.1/78148 2012-01-01T00:00:00Z http://hdl.handle.net/1721.1/72848 A priori and on-line route optimization for unmanned underwater vehicles Crimmel, Brian A The U.S. military considers Unmanned Underwater Vehicles (UUVs) a critical component of the future for two primary reasons - they are effective force multipliers and a significant risk-reducing agent. As the military's technology improves and UUVs become a reliable mission asset, the vehicle's ability to make intelligent decisions will be crucial to future operations. The thesis develops various algorithms to solve the UUV Mission-Planning Problem (UUVMPP), where the UUV must choose which tasks to perform in which sequence in a stochastic mission environment. The objective is to find the most profitable way to execute tasks with restrictions of total mission time, energy, time-restricted areas, and weather conditions. Since the UUV accumulates navigation error over time while maneuvering underwater, the UUV must occasionally halt operations to re-orient itself via a navigation fix. While a navigation fix takes time and increases the likelihood of exposing the vehicle's position to potential adversaries, a reduction in navigation error allows the UUV to perform tasks and navigate with a greater amount of certainty. The algorithms presented in this thesis successfully incorporate navigation fixes into the mission-planning process. The thesis considers Mixed-Integer Programming, Exact Dynamic Programming, and an Approximate Dynamic Programming technique known as Rollout to determine the optimal a priori route that meets operational constraints with a specified probability. The thesis then shows how these formulations can solve and re-solve the UUVMPP on-line. In particular, the Rollout Algorithm finds task route solutions on average 96% of the optimal solution a priori and 98% of the optimal solution on-line compared to exact algorithms; with a significant reduction in computation run time, the Rollout Algorithm permits the solving of increasingly complex mission scenarios. Thesis (S.M.)--Massachusetts Institute of Technology, Sloan School of Management, Operations Research Center, 2012.; Cataloged from PDF version of thesis.; Includes bibliographical references (p. 155-156). Sun, 01 Jan 2012 00:00:00 GMT http://hdl.handle.net/1721.1/72848 2012-01-01T00:00:00Z http://hdl.handle.net/1721.1/72647 Modeling social response to the spread of an infectious disease Evans, Jane A. (Jane Amanda) With the globalization of culture and economic trade, it is increasingly important not only to detect outbreaks of infectious disease early, but also to anticipate the social response to the disease. In this thesis, we use social network analysis and data mining methods to model negative social response (NSR), where a society demonstrates strain associated with a disease. Specifically, we apply real world biosurveillance data on over 11,000 initial events to: 1) describe how negative social response spreads within an outbreak, and 2) analytically predict negative social response to an outbreak. In the first approach, we developed a meta-model that describes the interrelated spread of disease and NSR over a network. This model is based on both a susceptible-infective- recovered (SIR) epidemiology model and a social influence model. It accurately captured the collective behavior of a complex epidemic, providing insights on the volatility of social response. In the second approach, we introduced a multi-step joint methodology to improve the detection and prediction of rare NSR events. The methodology significantly reduced the incidence of false positives over a more conventional supervised learning model. We found that social response to the spread of an infectious disease is predictable, despite the seemingly random occurrence of these events. Together, both approaches offer a framework for expanding a society's critical biosurveillance capability. Thesis (S.M.)--Massachusetts Institute of Technology, Sloan School of Management, Operations Research Center, 2012.; This electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections.; Cataloged from student-submitted PDF version of thesis.; Includes bibliographical references (p. 85-88). Sun, 01 Jan 2012 00:00:00 GMT http://hdl.handle.net/1721.1/72647 2012-01-01T00:00:00Z http://hdl.handle.net/1721.1/72646 Survivable paths in multilayer networks Parandehgheibi, Marzieh We consider the problem of protection in multilayer networks. In single-layer net- works, a pair of disjoint paths can be used to provide protection for a source-destination pair. However, this approach cannot be directly applied to layered networks where disjoint paths may not always exist. In this thesis, we take a new approach which is based on finding a set of paths that may not be disjoint but together will survive any single physical link failure. First, we consider the problem of finding the minimum number of survivable paths. In particular, we focus on two versions of this problem: one where the length of a path is restricted, and the other where the number of paths sharing a fiber is restricted. We prove that in general, finding the minimum survivable path set is NP-hard, whereas both of the restricted versions of the problem can be solved in polynomial time. We formulate the problem as Integer Linear Programs (ILPs), and use these formulations to develop heuristics and approximation algorithms. Next, we consider the problem of finding a set of survivable paths that uses the minimum number of fibers. We show that this problem is NP-hard in general, and develop heuristics and approximation algorithms with provable approximation bounds. We also model the dependency of communication networks on the power grid as a layered network, and investigate the survivability of communication networks in this layered setting. Finally, we present simulation results comparing the different algorithms. Thesis (S.M.)--Massachusetts Institute of Technology, Sloan School of Management, Operations Research Center; and, (S.M.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 2012.; This electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections.; Cataloged from student-submitted PDF version of thesis.; Includes bibliographical references (p. 75-77). Sun, 01 Jan 2012 00:00:00 GMT http://hdl.handle.net/1721.1/72646 2012-01-01T00:00:00Z