Penn Engineering

The School of Engineering and Applied Science, established in 1852, is composed of six academic departments and numerous interdisciplinary centers, institutes, and laboratories. At Penn Engineering, we are preparing the next generation of innovative engineers, entrepreneurs and leaders. Our unique culture of cooperation and teamwork, emphasis on research, and dedicated faculty advisors who teach as well as mentor, provide the ideal environment for the intellectual growth and development of well-rounded global citizens.

Search results

Now showing 1 - 10 of 4573
  • Publication
    Generation, Estimation and Tracking of Faces
    (1998) DeCarlo, Douglas M
    This thesis describes techniques for the construction of face models for both computer graphics and computer vision applications. It also details model-based computer vision methods for extracting and combining data with the model. Our face models respect the measurements of populations described by face anthropometry studies. In computer graphics, the anthropometric measurements permit the automatic generation of varied geometric models of human faces. This is accomplished by producing a random set of face measurements generated according to anthropometric statistics. A face fitting these measurements is realized using variational modeling. In computer vision, anthropometric data biases face shape estimation towards more plausible individuals. Having such a detailed model encourages the use of model-based techniques—we use a physics-based deformable model framework. We derive and solve a dynamic system which accounts for edges in the image and incorporates optical flow as a motion constraint on the model. Our solution ensures this constraint remains satisfied when edge information is used, which helps prevent tracking drift. This method is extended using the residuals from the optical flow solution. The extracted structure of the model can be improved by determining small changes in the model that reduce this error residual. We present experiments in extracting the shape and motion of a face from image sequences which exhibit the generality of our technique, as well as provide validation.
  • Publication
    From Deployments Of Elder Care Service Robots To The Design Of Affordable Low-Complexity End-Effectors And Novel Manipulation Techniques
    (2021-01-01) Rodrigues Mucchiani, Caio Cesar Carlos
    This thesis proposes an investigation on both behavioral and technical aspects of human-robot interaction (HRI) in elder care settings, in view of an affordable platform capable of executing desired tasks. The behavioral investigation combines a qualitative study with focus groups and surveys from not only the elders’ standpoint, but also from the standpoint of healthcare professionals to investigate suitable tasks to be accomplished by a service robot in such environments. Through multiple deployments of various robot embodiments at actual elder care facilities (such as at a low-income Supportive Apartment Living, SAL, and Program of All-Inclusive Care, PACE Centers) and interaction with older adults, design guidelines are developed to improve on both interaction and usability aspects. This need assessment informed the technical investigation of this work, where we initially propose picking and placing objects using end-effectors without internal mobility (or zero degrees-of-freedom, DOF), considering both quasi-static (tipping and regrasping as in-hand manipulation) and dynamic approaches. Maximizing grasping versatility by allowing robots to grasp multiple objects sequentially using a single end-effector and actuator is also proposed. These novel manipulation techniques and end-effector designs focus on minimizing robot hardware usage and cost, while still performing complex tasks and complying with safety constraints imposed by the elder care facilities.
  • Publication
    Machine Learning For Robot Motion Planning
    (2021-01-01) Zhang, Clark June
    Robot motion planning is a field that encompasses many different problems and algorithms. From the traditional piano mover's problem to more complicated kinodynamic planning problems, motion planning requires a broad breadth of human expertise and time to design well functioning algorithms. A traditional motion planning pipeline consists of modeling a system and then designing a planner and planning heuristics. Each part of this pipeline can incorporate machine learning. Planners and planning heuristics can benefit from machine learned heuristics, while system modeling can benefit from model learning. Each aspect of the motion planning pipeline comes with trade offs between computational effort and human effort. This work explores algorithms that allow motion planning algorithms and frameworks to find a compromise between the two. First, a framework for learning heuristics for sampling-based planners is presented. The efficacy of the framework depends on human designed features and policy architecture. Next, a framework for learning system models is presented that incorporates human knowledge as constraints. The amount of human effort can be modulated by the quality of the constraints given. Lastly, semi-automatic constraint generation is explored to enable a larger range of trade-offs between human expert constraint generation and data driven constraint generation. We apply these techniques and show results in a variety of robotic systems.
  • Publication
    Computer-Aided Clinical Trials For Medical Devices
    (2021-01-01) Jang, Kuk Jin
    Life-critical medical devices require robust safety and efficacy to treat patient populations with potentially large patient heterogeneity. Today, the de facto standard for evaluating medical devices is the randomized controlled trial. However, even after years of device development many clinical trials fail. For example, in the Rhythm ID Goes Head to Head Trial (RIGHT) the risk for inappropriate therapy by implantable cardioverter defibrillators (ICDs) actually increased relative to control treatments. With recent advances in physiological modeling and devices incorporating more complex software components, population-level device outcomes can be obtained with scalable simulations. Consequently, there is a need for data-driven approaches to provide early insight prior to the trial, lowering the cost of trials using patient and device models, and quantifying the robustness of the outcome. This work presents a clinical trial modeling and statistical framework which utilizes simulation to improve the evaluation of medical device software, such as the algorithms in ICDs. First, a method for generating virtual cohorts using a physiological simulator is introduced. Next, we present our framework which combines virtual cohorts with real data to evaluate the efficacy and allows quantifying the uncertainty due to the use of simulation. Results predicting the outcome of RIGHT and improving statistical power while reducing the sample size are shown. Finally, we improve device performance with an approach using Bayesian optimization. Device performance can degrade when deployed to a general population despite success in clinical trials. Our approach improves the performance of the device with outcomes aligned with the MADIT-RIT clinical trial. This work provides a rigorous approach towards improving the development and evaluation of medical treatments.
  • Publication
    Optical Studies of the Structure and Dynamics of Opaque Colloids
    (1993) Kaplan, Peter D
    Three fundamental experiments on multi-particle interactions in hard-sphere systems are presented in this thesis. In the first experiment, a mixture of two monodisperse colloids with different particle diameters, but the same particle composition, is studied using diffusing-wave spectroscopy. We find that both the particle structure and the hydrodynamic interactions between large and small balls are well modeled by existing theories. These theories, however, do not comment on the observed diffusion of particles in complex solutions over distances larger than a particle diameter. Incidental observations during this experiment led to the second experiment, the study of the phase diagram of binary hard-sphere mixtures of colloidal particles. In the process of mapping this phase diagram, driven entirely by entropy, a novel surface phase was discovered. This surface phase is explained geometrically in terms of the entropy of the small particles in the suspension. In the final experiment, the diffuse model of photon transport is extended to create a new technique called Diffuse Transmission Spectroscopy (DTS) for exploiting the wave vector dependent structure of colloidal suspensions. DTS uses the wavelength dependent transmission of a slab of colloid to probe interparticle structure. The diffuse transmission spectrum is shown to be useful for measuring particle diameter and differences in interparticle potential in dense, undiluted systems, as well as for testing the validity of structural models.
  • Publication
    Nonlinear Optical Spectroscopy of Solid/Solid Interfaces
    (1992-12-01) Yeganeh, Mohsen S
    Three-wave mixing (3WM) spectroscopy is an exciting and relatively unexplored probe of buried solid interfaces. It possesses long penetration depths characteristic of most optical methods and intrinsic interface specificity characteristic of second-order optical processes. In this thesis we present frequency domain measurements of the ZnSe/GaAs(OOl) heterojunction by second-harmonic (SH) and sum-frequency (SF) generation. Our experiments reveal an unusual three-wave mixing resonance that arises as a result of virtual transitions between an interfacial quantum well state and the ZnSe valence band. The interfacial quantum well was brought about by interdiffusion of Zn (Ga) into GaAs (ZnSe) during sample growth. The observation introduces a new class of nonlinear optical phenomena at interfaces that can provide useful information about band profiles, diffusion and defects along the boundary of two semiconductors. We have found that this interfacial SH resonance is sensitive to a variety of structural phenomena. In essence any process that modifies the band profile near the junction will affect the strength of the resonance. We have observed the variation of interface SH spectra with respect to lattice strain relaxation and to surface reconstruction of the buried GaAs. In addition, using a newly developed photomodulationSHG (PSHG) technique, we have exploited this sensitivity to determine the nature and relative density of interface charge traps as a function of substrate surface reconstruction. The PSHG method was also used to study free charge trapping mechanisms at ZnSe/GaAs(OOl) heterointerfaces. Our measurements determined that the interfacial trap-centers are mainly hole-traps with lifetime of 35 sec. In the course of carrying out these experiments we also observed interference in reflected second harmonic generation from two adjoined nonlinear slabs. A theory for the phenomena was presented and was used to understand our experimental results with ZnSe/GaAs(OOl) heterostructures. This interference phenomena was introduced as a new methodology to measure the second-order susceptibility of thin overlayer materials.
  • Publication
    Imu-Based State Estimation And Control Of Quadrotors Exploiting Aerodynamic Effects
    (2019-01-01) Svacha, James Baird
    Quadrotors and multirotors in general are common in many inspection and surveillance applications. For these applications, visual-inertial odometry is a common way to localize the vehicles and observe the environment. However, unlike with wheeled mobile robots, quadrotor localization algorithms often do not use knowledge of the control inputs and the full vehicle dynamics as a process model for localization. Rather, they use kinematic models, with the IMU providing acceleration and angular velocity. One of the reasons for avoiding the use of dynamics is that, until recently quadrotor aerodynamic effects have not been considered in the literature and hence the dynamic models for quadrotors have been less accurate than those for wheeled mobile robots. The main aerodynamic terms that are significant are first-order effects that are linear in velocity and angular velocity. They are predominantly caused by aerodynamic interaction with the spinning propellers. This work investigates the models for such effects, as well as what can be gained if such aerodynamic effects are incorporated into the dynamic model and the full dynamics are used for state estimation. We develop novel IMU-based filters, the end results of which are used to estimate the wind velocity of the quadrotor or, indoors, when the ambient wind is zero, the velocity of the quadrotor. In addition, these filters estimate the many aerodynamic parameters in the model online. They may also be used to estimate sensor biases and inertial parameters. We demonstrate the effectiveness of these filters through experiments. We also present nonlinear observability analyses that theoretically determine the observability properties of the systems.
  • Publication
    Multi-band Oversampled Noise Shaping Analog to Digital Conversion
    (1996) Aziz, Pervez M
    Oversampled noise shaping analog to digital (A/D) converters, which are commonly known as delta-sigma (ΔΣ) converters, have the ability to convert relatively low bandwidth signals with very high resolution. Such converters achieve their high resolution by oversampling, as well as processing the signal and quantization noise with different transfer functions. The signal transfer function (STF) is typically a delay over the signal band while the noise transfer function (NTF) is designed to attenuate quantization noise in the signal band. A side effect of the NTF is an amplification of the noise outside the signal band. Thus, a digital filter subsequently attenuates the out-of-band quantization noise. The focus of this thesis is the investigation of ΔΣ architectures that increase the bandwidth where high resolution conversion can be achieved. It uses parallel architectures exploiting frequency or time slicing to meet this objective. Frequency slicing involves quantizing different portions of the signal frequency spectrum using several quantizers in parallel and then combining the results of the quantizers to form an overall result. Time slicing involves quantizing various groups of time domain signal samples with different quantizers in parallel and then combining the results of the quantizers to form an overall output. Several interesting observations can be made from this general perspective of frequency and time slicing. Although the representation of a signal are completely equivalent in time or frequency, the thesis shows that this is not the case for known frequency and time sliced A/D architectures. The performance of such systems under ideal conditions are compared for PCM as well as for ΔΣ A/D converters. A multi-band frequency sliced architecture for delta-sigma conversion is proposed and its performance is included in the above comparison. The architecture uses modulators which realize different NTFs for different portions of the signal band. Each band is converted in parallel. A bank of FIR filters attenuates the out of-band noise for each band and achieves perfect reconstruction of the signal component. A design procedure is provided for the design of the filter bank with reduced computational complexity. The use of complex NTFs in the multi-band ΔΣ architecture is also proposed. The peformance of real and complex NTFs is compared. Performance evaluations are made for ideal systems as well as systems suffering from circuit implementation imperfections such as finite opamp gain and mismatched capacitor ratios.
  • Publication
    Building Parameterized Action Representations From Observation
    (2000-01-01) Bindiganavale, Ramamani
    Virtual worlds may be inhabited by intelligent agents who interact by performing various simple and complex actions. If the agents are human-like (embodied), their actions may be generated from motion capture or procedural animation. In this thesis, we introduce the CaPAR interactive system which combines both these approaches to generate agent-size neutral representations of actions within a framework called Parameterized Action Representation (PAR). Just as a person may learn a new complex physical task by observing another person doing it, our system observes a single trial of a human performing some complex task that involves interaction with self or other objects in the environment and automatically generates semantically rich information about the action. This information can be used to generate similar constrained motions for agents of different sizes. Human movement is captured by electromagnetic sensors. By computing motion zerocrossings and geometric spatial proximities, we isolate significant events, abstract both spatial and visual constraints from an agent's action, and segment a given complex action into several simpler subactions. We analyze each independently and build individual PARs for them. Several PARs can be combined into one complex PAR representing the original activity. Within each motion segment, semantic and style information is extracted. The style information is used to generate the same constrained motion in other differently sized virtual agents by copying the end-effector velocity profile, by following a similar end-effector trajectory, or by scaling and mapping force interactions between the agent and an object. The semantic information is stored in a PAR. The extracted style and constraint information is stored in the corresponding agent and object models.
  • Publication
    Parsing With Lexicalized Tree Adjoining Grammar
    (1990-02-01) Schabes, Yves; Joshi, Aravind K
    Most current linguistic theories give lexical accounts of several phenomena that used to be considered purely syntactic. The information put in the lexicon is thereby increased in both amount and complexity: see, for example, lexical rules in LFG (Kaplan and Bresnan, 1983), GPSG (Gazdar, Klein, Pullum and Sag, 1985), HPSG (Pollard and Sag, 1987), Combinatory Categorial Grammars (Steedman, 1987), Karttunen's version of Categorial Grammar (Karttunen 1986, 1988), some versions of GB theory (Chomsky 1981), and Lexicon-Grammars (Gross 1984). We would like to take into account this fact while defining a formalism. We therefore explore the view that syntactical rules are not separated from lexical items. We say that a grammar is lexicalized (Schabes, AbeilK and Joshi, 1988) if it consists of: (1) a finite set of structures each associated with lexical items; each lexical item will be called the anchor of the corresponding structure; the structures define the domain of locality over which constraints are specified; (2) an operation or operations for composing the structures. The notion of anchor is closely related to the word associated with a functor-argument category in Categorial Grammars. Categorial Grammar (as used for example by Steedman, 1987) are 'lexicalized' according to our definition since each basic category has a lexical item associated with it.