Made possible through the support of the National Science Foundation through a REU grant and with the support of Microsoft, Inc.

PERFORMANCE OPTIMIZATION OF
MILLIMETR WAVE LENSING USING METAMATERIAL CONCEPT
by Amber Sallerson
Abstract
Full Paper
Slides

A DISTORTION-MINIMIZING MICRO-MIRROR ARRAY WITH MICROACTUATORS FOR WIDE-ANGLE VIEWING
by Christopher Bremer
Abstract
Full Paper
Slides

A 3-D DISTRIBUTED MOBILE SENSOR NETWORK
by Yao Hua Ooi
Abstract
Full Paper
Slides

DISTRIBUTED NETWORK COMMUNICATION FOR AN OLFACTORY ROBOT
by Jiong Shen
Abstract
Full paper
Slides

ELECTROKINETICS OF MICROPARTICLES USING AC DIELECTROPHORESIS
by John Zelena
Abstract
Full paper
Slides

THE SMART CHAIR PROJECT: REDUCING CODING FOR THE SMART CHAIR PROGRAMMERS
by Catherine Lachance 
Abstract
Full paper
Slides

DESIGN OF AN ARTIFICIAL COCHLEA USING DIGITAL FILTERS ON A FIELD-PROGRAMMABLE GATE ARRAY
by Aslan Ettehadieh
Abstract
Full paper
Slides

MARS: MULTIPLE AUTONOMOUS ROBOTS
by Kamela Watson
Abstract
Full paper
Slides

ASSEMBLY AND CHARACTERIZATION OF NANOELECTROMECHANICAL SYSTEMS
by Cynthia Moreno
Abstract
Full Paper
Slides

AN ELECTRIC POWER SYSTEM FOR AN AUTONOMOUS BLIMP
by Frederik Heger
Abstract
Full paper
Slides

 
 
 
 

EFFECT OF COHERENT MOVING STIMULUS ON THE VISUAL EVOKED POTENTIAL (CMVEP)
Adrian Lau (Electrical Engineering) – University of Pennsylvania
Advisors: Dr. Nader Engheta and Dr. Edward Pugh

This paper describes research on visually evoked potential by motion stimulus through the use of electroencephalography. The purpose was to investigate the human perception of coherent motion and incoherent motion by conducting experiments to determine the response to a coherent moving stimulus. The response was further verified by varying the stimulus parameters and using several different approaches, based on current understanding of the motion detection mechanism in the neural system. Positive results from the experiment provide further evidence for the proposed motion detection mechanism. Based on the results of the experiments, the properties of the coherent motion stimulus evoked potential are characterized.

Because of impracticalities resulting from their much larger size, lensing systems in a millimeter wave band cannot be dealt with in the same manner as optical systems. One technique for implementing this lensing is to transform the phase of the incident field, using metamaterials such as frequency-selective surfaces (FSSs). In this project, the FSSs known as the gangbuster surface (GS) is used to achieve the lensing effect.

We consider the GS above a perfectly conducting ground plane. Incident waves reflect totally, and the phase of the reflected waves depends on characteristics such as the type and period of the GS, the lengths of the wires, and the separation between the GS and the ground plane. Any change in the various parameters used to design the lens results in deterioration in the profile field intensity and the diffraction efficiency, among other unwanted effects.

The goal is to optimize the lens for changes in the angle of incidence such that the intensity of the electric field and the resolution at the focal plane are both optimized. To accomplish this optimization, differences in phase distribution for various angles are minimized by using an existing genetic algorithm to modify the lengths of wires on the GS. The incorporation of an objective function into the genetic algorithm returns the optimal length of wire needed to achieve the desired phase distribution. Through the use of these different wire lengths, a GS can be constructed for a lensing system in the millimeter wave band that is stable for differing angles of incidence.

Mirrors of numerous shapes, including spherical and paraboloidal mirrors, have been employed for many different commercial and industrial uses, despite their tendencies to distort and warp images. When the object in consideration is planar and oriented normal to the optical axis of the mirror, the amount of distortion introduced by a paraboloidal mirror is less than the distortion introduced by a spherical mirror. Therefore, there must exist an optimal mirror shape that minimizes distortion of planar objects normal to the optical axis. This objective of this study was to fabricate a versatile device capable of simulating numerous mirror shapes to allow for determination of optimal, distortion- minimizing mirrors. Using an Excimer Laser, millimeter-scale mirror mechanisms capable of rotation about two axes were created. Electromagnetic actuation was tested and experimented with in order to determine its feasibility as a means of mirror manipulation. Piezoelectric actuation was also investigated. Although no actuation was achieved, the numerous attempts provided useful information as to why no motion occurred. This paper will delineate the findings of these actuation experiments and provide the groundwork for future research and experimentation. 

Biological organisms employ binocular visual and binaural hearing systems, as well as movement, to gather sensory information from multiple viewpoints for accurate sensory perception. In contrast, artificial sensory systems typically use either a multitude of sensors in a static array, or employ motion from a single mobile robotic platform. 

The goal of this project is to build a small prototype of an adaptive, distributed network consisting of small, modular sensors and actuating components that will accurately position sensors at multiple 3-dimensional spatial locations and yield sensory information from multiple viewpoints. A working prototype of a single sensor node system was built, using a simple motorized spool design and a Motorola HC11 microcontroller. Control of the system was implemented in C and assembly to position a single sensor node in a 2-dimensional space. A user interface, allowing input via infra-red remote control was designed. 

The olfactory robot combines a nose sensor with robotics to provide automated information gathering without human labor. To achieve this goal, however, the information the robot gathers must be transferred to a computer to be analyzed. One way to transfer the data is through a wireless network that provides mobility as well as high speed. For this project, a strong-arm processor machine, Netwinder, is used to send information from the olfactory robot. Then, the sending and receiving C++ object-oriented codes were designed to use UDP multicast send and receive with RTP-specific packets to form a distributed network. Multicast network provides an easy way to distribute information from one sender to multiple receivers by using a group IP address. Then to further improve the speed of the data transfers, shared memory is implemented to control the sending and receiving flow so that no bottlenecks occur in between. Moreover, the shared memory provides a way for the receiver computer to distribute its information among its internal processes. The result achieved was the continuous streaming of the JPEG image frames from the source to any computers listening on the multicast channel. The receiver could receive and pipe the images to multiple processes through the shared memory, creating a large two-level distributed network.

Meso-scale (approximately 10 µm to 1000 µm) systems have a variety of applications, including use in medical and biological fields, automotives, and space technology. An electro-kinetic device for manipulating and moving biological micro-particles is the focus of this paper. Rectilinear motion of the particles is caused by using traveling wave dielectrophoresis. The theory behind dielectrophoresis (DC and AC), design, forces, fabrication methods, and results, are discussed.

The Smart Chair is a wheelchair that allows a user to navigate and communicate solely through an onboard computer.  The research described in this paper dealt mainly with organizing the program’s interface and facilitating the creation of the pages displayed on the Smart Chair.  These innovations, including writing auxiliary functions and combining similar variables into larger data structures, make the wheelchair more user-friendly and make its code easier to read.  Work was also started this summer to approach page display in a new way, so that code needed to generate a page was created dynamically at runtime, and thus one function would suffice to generate all pages, instead of one function per page.  Furthermore, by creating a graphical user interface that generates the code needed to display each page, this research cut the amount of coding needed to run the program by more than forty-two percent, and thus minimized the total time needed to program the Smart Chair.  Because of this newfound convenience for the programmer, the Smart Chair program can be expanded more easily, with more options for the user.

Traditional methods of speech recognition have very limited complexity and impose considerable grammar constraints.  Today’s systems have critical problems understanding different voices and do not have robust vocabularies.  This paper describes research on a biological method for speech recognition that models an artificial cochlea using digital filters.  The specific part of the cochlea of interest here, the basilar membrane acts as a collection of bandpass filters that will be mimicked to develop an artificial digital cochlea. 

This digital cochlea will be implemented on a single Xilinx field programmable gate array (FPGA).  The FPGA chip is on a board that contains both an analog to digital (A/D) converter and a digital to analog (D/A) converter.  The A/D converter converts the speech signal into a digital representation; the D/A converter converts the digital signal back into its original analog form.  The board also includes a lowpass anti-aliasing filter to reduce the noise and keep only the range of human speech frequencies approximately 100 to 3500 Hertz.  The range of frequencies will be divided into 16 bands.  The FPGA chip will contain 16 programmable bandpass filters to split up the speech signal into separate frequency components that can be used to determine phonemes, the simplest unit of speech.  Phoneme-level recognition will improve the speed and accuracy of speech recognition.

The final product is expected to be cost efficient and to be implemented on a single chip.  The digital cochlea will be used in conjunction with a neural network that will extract features and phonemes from speech signals.

Robotics is becoming an increasingly important field in engineering research. The MARS (Multiple Autonomous Robots) team at the University of Pennsylvania fabricates small-scale, car-like robots. These robots are fitted with 12 infrared sensors and an omni directional camera for sensing as well as an onboard laptop computer with a wireless network. The robots can perform a variety of functions and tasks that, although seemingly simple individually, can be combined and manipulated to allow the robots to perform more complicated tasks.

Over the course of the summer, I have participated heavily in the upgrade of the hardware on the robots. Participating in efforts on calibration, computer casing, and robot chassis, I have helped to make improvements, introduce new ideas, and documenting all of my summer efforts. The following will discuss the overall robot structure and then continue to explain my findings and efforts to the MARS project over the course of the summer. 

Nanoelectromechanical systems, or NEMS, are characterized by minute dimensions that are relevant for the function of the devices. Using varied methods, many studies have sought to integrate nanostructures with functional circuitry. Some studies have focused on alternative methods that rely on the assembly of nanometer scale particles such as an electric-field assembly technique. The research reported here focused on the acquisition and development of a new experimental system that has enabled the control of individual nanostructures onto their predefined sites on the silicon chips. An electric-field assisted assembly technique was used to manipulate dielectric particles suspended in a medium between two electrodes 1 to 3 microns apart, defined lithographically on an SiO2 substrate. Assembly experiments were first conducted in which a few drops of the medium were dispensed onto the sample and a 30 V AC signal fed at 1 kHz for varying lengths of time to the silicon chip that contains a series of prefabricated buried inter-digitated electrodes. 

Keywords: nanoelectromechanical, electric-field assembly

This paper describes the design and implementation of an experimental electric power system to replace a gasoline system of a 30-feet unmanned autonomous blimp. The project was motivated by limitations and inefficiencies of the gasoline system. Autonomous flight requires extended duration flight and maneuvering capabilities that can be realized with an electric power system. In order to fully exploit the advantages of electric flight, the gondola housing the power and control systems was redesigned for minimum weight as a carbon fiber frame structure. With the optimal motor-propeller combination for the chosen motors, electric flight was possible at less than half of the power compared to the gasoline system indoors for 30 minutes. Increased maneuverability gained from the ability to stop, restart and reverse the motors in flight was an important step towards controlling the blimp with an onboard computer. The weight reduction of the gondola effectively increased the payload capacity, allowing for neutrally buoyant flight and additional sensor and control equipment. The electric power system significantly increased the blimp’s flight capabilities and provided the foundation for future work towards a fully autonomous blimp capable of operating alone or together with other robots in a multi-robot environment.