On robots swarm dispersion, task scheduling in distributed systems, and Web-based short period style analysis.

摘要

In the first section of the dissertation we examine some problems arising in swarm robotics. Swarm robotics is one of the most exciting paradigms in the fields of robotics and artificial intelligence. The goal is to perform relatively complex tasks by employing a large number of small, very simple, autonomous, and inexpensive robots. In this work, we explore the process of dispersing a swarm of robots from an initially densely packed configuration. Several dispersing protocols are proposed and tested in a simulation environment.; In the second section of the dissertation we study a problem arising in distributed computing. With the modern advancement of networking and decentralization of computing resources, it has become plausible to perform computationally intensive jobs using a network of distributed machines. We study a “work stealing” algorithm, called the Enhanced Cilk Scheduler, for task scheduling in distributed systems. We perform simulation experiments comparing our algorithm to an algorithm based on a centrally managed greedy scheduler. Our analysis shows that the Enhanced Cilk Scheduler is a viable distributed system scheduler.; In the third section of the dissertation we develop a web-based financial advising tool. We give a web-based style analysis tool together with its core solver, which requires a fast implementation of a constrained quadratic programming algorithm. Style analysis is a very useful technique used in the investment process. It provides insight into the composition of the portfolio with respect to asset classes based solely on returns. The solver is used in the process of finding a custom-tailored investment portfolio, in analyzing the style of a portfolio composed of an unknown set of securities, and in other financial applications. The optimizer, written in the Java programming language, uses a gradient method that is simple yet fast in finding an approximate solution to the problem, up to a specified precision.