This paper describes an intuitive way of defining geometry design variables for solving structural topology optimization problems using a genetic algorithm (GA). The geometry representation scheme works by defining a skeleton that represents the underlying topology/connectivity of the continuum structure. As the effectiveness of any GA is highly dependent on the chromosome encoding of the design variables, the encoding used here is a directed graph which reflects this underlying topology so that the genetic crossover and mutation operators of the GA can recombine and preserve any desirable geometric characteristics through succeeding generations of the evolutionary process. The overall optimization procedure is tested by solving a simulated topology optimization problem in which a 'target' geometry is pre-defined with the aim of having the design solutions converge towards this target shape. The procedure is also applied to design a straight-line compliant mechanism : a large displacement flexural structure that generates a vertical straight line path at some point when given a horizontal straight line input displacement at another point. / Singapore-MIT Alliance (SMA)

This paper describes the initial results of a project to create a self-supervised algorithm for learning object segmentation from video data. Developmental psychology and computational experience have demonstrated that the motion segmentation of objects is a simpler, more primitive process than the detection of object boundaries by static image cues. Therefore, motion information provides a plausible supervision signal for learning the static boundary detection task and for evaluating performance on a test set. A video camera and previously developed background subtraction algorithms can automatically produce a large database of motion-segmented images for minimal cost. The purpose of this work is to use the information in such a database to learn how to detect the object boundaries in novel images using static information, such as color, texture, and shape. / Singapore-MIT Alliance (SMA)

The Weapon Target Assignment (WTA) problem is a fundamental problem arising in defense-related applications of operations research. This problem consists of optimally assigning n weapons to m targets so that the total expected survival value of the targets after all the engagements is minimum. The WTA problem can be formulated as a nonlinear integer programming problem and is known to be NP-complete. There do not exist any exact methods for the WTA problem which can solve even small size problems (for example, with 20 weapons and 20 targets). Though several heuristic methods have been proposed to solve the WTA problem, due to the absence of exact methods, no estimates are available on the quality of solutions produced by such heuristics. In this paper, we suggest linear programming, integer programming, and network flow based lower bounding methods using which we obtain several branch and bound algorithms for the WTA problem. We also propose a network flow based construction heuristic and a very large-scale neighborhood (VLSN) search algorithm. We present computational results of our algorithms which indicate that we can solve moderately large size instances (up to 80 weapons and 80 targets) of the WTA problem optimally and obtain almost optimal solutions of fairly large instances (up to 200 weapons and 200 targets) within a few seconds

This paper develops a potential reduction algorithm for solving a linear-programming problem directly from a "warm start" initial point that is neither feasible nor optimal. The algorithm is of an "interior point" variety that seeks to reduce a single potential function which simultaneously coerces feasibility improvement (Phase I) and objective value improvement (Phase II). The key feature of the algorithm is the ability to specify beforehand the desired balance between infeasibility and nonoptimality in the following sense. Given a prespecified balancing parameter /3 > 0, the algorithm maintains the following Phase I - Phase II "/3-balancing constraint" throughout: (cTx- Z*) < /3TX, where cTx is the objective function, z* is the (unknown) optimal objective value of the linear program, and Tx measures the infeasibility of the current iterate x. This balancing constraint can be used to either emphasize rapid attainment of feasibility (set large) at the possible expense of good objective function values or to emphasize rapid attainment of good objective values (set /3 small) at the possible expense of a lower infeasibility gap. The algorithm exhibits the following advantageous features: (i) the iterate solutions monotonically decrease the infeasibility measure, (ii) the iterate solutions satisy the /3-balancing constraint, (iii) the iterate solutions achieve constant improvement in both Phase I and Phase II in O(n) iterations, (iv) there is always a possibility of finite termination of the Phase I problem, and (v) the algorithm is amenable to acceleration via linesearch of the potential function.

I present an algorithm which allows two agents to generate a simple language based only on observations of a shared environment. Vocabulary and roles for the language are learned in linear time. Communication is robust and degrades gradually as complexity increases. Dissimilar modes of experience will lead to a shared kernel vocabulary.

"Expectation-Maximization'' (EM) algorithm and gradient-based approaches for maximum likelihood learning of finite Gaussian mixtures. We show that the EM step in parameter space is obtained from the gradient via a projection matrix $P$, and we provide an explicit expression for the matrix. We then analyze the convergence of EM in terms of special properties of $P$ and provide new results analyzing the effect that $P$ has on the likelihood surface. Based on these mathematical results, we present a comparative discussion of the advantages and disadvantages of EM and other algorithms for the learning of Gaussian mixture models.

This work is aimed towards the development of an artificially intelligent search algorithm used in conjunction with an Auto Associative Neural Network (AANN) to help locate and reconstruct faulty sensor inputs in control systems. The AANN can be trained to detect when sensors go faulty but the problem of locating the faulty sensor still remains. The search algorithm aids the AANN to help locate the faulty sensors and reconstruct their actual values. The algorithm uses domain specific heuristics based on the inherent behavior of the AANN to achieve its task. Common sensor errors such as drift, shift and random errors and the algorithms response to them have been studied. The issue of noise has also been investigated. These areas cover the first part of this work. The second part focuses on the development of a web interface that implements and displays the working of the algorithm. The interface allows any client on the World Wide Web to connect to the engineering software called MATLAB. The client can then simulate a drift, shift or random error using the graphical user interface and observe the response of the algorithm.

In-process Optical Coordinate Measuring Machine (OCMM) offers the potential of diagnosing in a timely manner variation sources that are responsible for product quality defects. Such a sensor system can help manufacturers improve product quality and reduce process downtime. Effective use of sensory data in diagnosing variation sources depends on the optimal design of a sensor system, which is often known as the problem of sensor placements. This thesis addresses coordinate sensor placement in diagnosing dimensional variation sources in assembly processes. Sensitivity indices of detecting process mean and variance components are defined as the design criteria and are derived in terms of process layout and sensor deployment information. Exchange algorithms, originally developed in the research of optimal experiment deign, are employed and revised to maximize the detection sensitivity. A sort-and-cut procedure is used, which remarkably improve the algorithm efficiency of the current exchange routine. The resulting optimal sensor layouts and its implications are illustrated in the specific context of a panel assembly process.

The rapid emergence of distributed fuel cell powered generations (DFPGs) operating in parallel with utility has brought a number of technical concerns as more DFPGs are connected to utility grid. One of the most challenging problems is known as islanding phenomenon. This situation occurs when a network is disconnected from utility grid and is energized by local DFPGs. It can possibly result in injury to utility personnel arriving to service isolated feeders, equipment damage, and system malfunction. In response to the concern, this dissertation aims to develop a robust anti-islanding algorithm for utility interconnection of DFPGs. In the first part, digital signal processor (DSP) controlled power electronic converters for utility interconnection of DFPGs are developed. Current control in a direct-quadrature (dq) synchronous frame is proposed. The real and reactive power is controlled by regulating inverter currents. The proposed digital current control in a synchronous frame significantly enhances the performance of DFPGs. In the second part, the robust anti-islanding algorithm for utility interconnection of a DFPG is developed. The power control algorithm is proposed based on analysis of a real and reactive power mismatch. It continuously perturbs (±5%) the reactive power supplied by the DFPG while monitoring the voltage and frequency. If islanding were to occur, a measurable frequency deviation would take place, upon which the real power of the DFPG is further reduced to 80%; a drop in voltage positively confirms islanding. This method is shown to be robust and reliable. In the third part, an improved anti-islanding algorithm for utility interconnection of multiple DFPGs is presented. The cross correlation method is proposed and implemented in conjunction with the power control algorithm. It calculates the cross correlation index of a rate of change of the frequency deviation and (±5%) the reactive power. If this index increases above 50%, the chance of islanding is high. The algorithm initiates (±10%) the reactive power and continues to calculate the correlation index. If the index exceeds 80%, islanding is now confirmed. The proposed method is robust and capable of detecting islanding in the presence of several DFPGs independently operating. Analysis, simulation and experimental results are presented and discussed.

This thesis describes the use of an ultrasonic goal-seeking algorithm while using ionic polymer metal composite (IPMC), an electroactive polymer, as the actuator to drive a vessel towards a goal. The signal transmitting and receiving circuits as well as the goal seeking algorithm are described in detail. Two test vessels were created; one was a larger vessel that contained all necessary components for autonomy. The second was a smaller vessel that contained only the sensors and IPMC strips, and all power and signals were transmitted via an umbilical cord. To increase the propulsive efforts of the second, smaller vessel, fins were added to the IPMC strips, increasing the surface area over 700%, determined to yield a 22-fold force increase. After extensive testing, it was found that the three IPMC strips, used as oscillating fins, could not generate enough propulsion to move either vessel, with or without fins. With the addition of fins, the oscillating frequency was reduced from 0.86-Hz to 0.25-Hz. However, the goal-seeking algorithm was successful in guiding the vessel towards the target, an ultrasonic transmitter. When moved manually according to the instructions given by the algorithm, the vessel successfully reached the goal. Using assumptions based on prior experiments regarding the speed of an IPMC propelled vessel, the trial in which the goal was to the left of the axis required 18.2% more time to arrive at the goal than the trial in which the goal was to the right. This significant difference is due to the goal-seeking algorithm’s means to acquire the strongest signal. After the research had concluded and the propulsors failed to yield desired results, many factors were considered to rationalize the observations. The operating frequency was reduced, and it was found that, by the impulse-momentum theorem, that the propulsive force was reduced proportionally. The literature surveyed addressed undulatory motion, which produces constant propulsive force, not oscillatory, which yields intermittent propulsive force. These reasons among others were produced to rationalize the results and prove the cause of negative results was inherent to the actuators themselves. All rational options have been considered to yield positive results.