Judgments about personality based on facial appearance are strong effectors in social decision making, and are known to have impact on areas from presidential elections to jury decisions. Recent work has shown that it is possible to predict perception of memorability, trustworthiness, intelligence and other attributes in human face images. The most successful of these approaches require face images expertly annotated with key facial landmarks. We demonstrate a Convolutional Neural Network (CNN) model that is able to perform the same task without the need for landmark features, thereby greatly increasing efficiency. The model has high accuracy, surpassing human-level performance in some cases. Furthermore, we use a deconvolutional approach to visualize important features for perception of 22 attributes and demonstrate a new method for separately visualizing positive and negative features.

This paper presents a low-complexity, novel approach to wireless sensor network (WSN) assisted autonomous mobile robot (AMR) navigation. The goal is to have an AMR navigate to a target location using only the information inherent to WSNs, i.e., topology of the WSN and received signal strength (RSS) information, while executing an efficient navigation path. Here, the AMR has neither the location information for the WSN, nor any sophisticated ranging equipment for prior mapping. Two schemes are proposed utilizing particle filtering based bearing estimation with RSS values obtained from directional antennas. Real-world experiments demonstrate the effectiveness of the proposed schemes. In the basic node-to-node navigation scheme, the bearing-only particle filtering reduces trajectory length by 11.7% (indoors) and 15% (outdoors), when compared to using raw bearing measurements. The advanced scheme further reduces the trajectory length by 22.8% (indoors) and 19.8% (outdoors), as compared to the basic scheme. The mechanisms exploit the low-cost, low-complexity advantages of the WSNs to provide an effective method for map-less and ranging-less navigation.

This paper extends the novel research for event localization and target-directed navigation using a deployed wireless sensor network (WSN) [4]. The goal is to have an autonomous mobile robot (AMR) navigate to a target-location by: (i) producing an artificial magnitude distribution within the WSN-covered region, and (ii) having the AMR use the pseudo-gradient from the interpolated distribution in its neighborhood, as it moves towards the target location. Implicit surfaces are used to interpolate the artificial distribution. This scheme only uses the topology of the WSN and received signal strength (RSS) to estimate an efficient navigation path for the AMR. Here, the AMR does not require global coordinates for the region, as it relies on local, neighborhood information alone to navigate. The performance of the scheme is analyzed with hardware experiments and in simulation, using a variety of node-densities and with increasing levels of noise to ensure robustness.

Venous leg ulcers remain a major problem in the United States, with spending reaching more than $1 billion annually. Current treatment options for this condition center around the use of compression therapy delivered by bandages, medical-grade stockings, or pneumatic compression devices. While these forms of therapy can produce dramatic improvements, cost and patient compliance remain an issue. In parallel with this need, wearable, wireless health monitoring systems have recently emerged as a low-cost solution for management of chronic health conditions. To this end, researchers at the Center for Robotics and Intelligent Machines (North Carolina State University) and the Carolon Company (Rural Hall, NC) have proposed an integrated sensing and therapeutic compression module. This article will review technologies related to the design of such a device, as well as provide direction for future research.

Blastocyst microinjections are routinely involved in the process of creating genetically modified mice for biomedical research, but their efficiency is highly dependent on the skills of the operators. As a consequence, much time and resources are required for training microinjection personnel. This situation has been aggravated by the rapid growth of genetic research, which has increased the demand for mutant animals. Therefore, increased productivity and efficiency in this area are highly desired. Here, we pursue these goals through the automation of a previously developed teleoperated blastocyst microinjection system. This included the design of a new system setup to facilitate automation, the definition of rules for automatic microinjections, the implementation of video processing algorithms to extract feedback information from microscope images, and the creation of control algorithms for process automation. Experimentation conducted with this new system and operator assistance during the cells delivery phase demonstrated a 75% microinjection success rate. In addition, implantation of the successfully injected blastocysts resulted in a 53% birth rate and a 20% yield of chimeras. These results proved that the developed system was capable of automatic blastocyst penetration and retraction, demonstrating the success of major steps toward full process automation.

In this paper, two versions of fabric-based active electrodes are presented to provide a wearable solution for ECG monitoring clothing. The first version of active electrode involved direct attachment of surface-mountable components to a textile screen-printed circuit using polymer thick film techniques. The second version involved attaching a much smaller, thinner, and less obtrusive interposer containing the active electrode circuitry to a simplified textile circuit. These designs explored techniques for electronic textile interconnection, chip attachment to textiles, and packaging of circuits on textiles for durability. The results from ECG tests indicate that the performance of each active electrode is comparable to commercial Ag/AgCl electrodes. The interposer-based active electrodes survived a five-cycle washing test while maintaining good signal integrity.

Blastocyst microinjections are routinely involved in the process of creating genetically modified mice for biomedical research, but their efficiency is typically poor and highly dependent on the skills of the operators. Here, the goal of increasing the consistency and efficiency rates of these operations was pursued through robotics and automation. Initially, a teleoperated microinjection system was designed and implemented. Then, this system was evaluated under the control of a microinjection expert. The evaluation consisted of a series of blastocyst microinjections, which are presented and analyzed in this paper. The knowledge gained from these experiments was then used to: 1) create a new system setup more appropriate for automating consecutive blastocyst microinjections; 2) derive rules to guide automatic microinjections; and 3) define future research directions and system upgrades to further increase the system performance. Finally, experiments with the encoded algorithm for automatic blastocyst microinjections attested the validity of the derived rules by demonstrating a 75% success rate. The good quality of the microinjections was demonstrated by the 53% birth rate and the 20% yield of chimeras obtained from the implanted blastocysts.

Many recent deployments of environmental sensor networks have focused on obtaining measurements across large and inhospitable areas. With increasing scale it becomes impractical to deploy or maintain such systems by hand. This paper evaluates large scale network disconnectivity and highlights the underlying issues related to the environment and node characteristics. Furthermore, it examines how a low cost and adaptive method of robotic repair can be applied to large area networks using received signal strength measurements for simple navigation and placement.

This paper presents results related to our latest semi-automated blastocyst microinjection system. Here, the improvements made to the microinjection system are described and evaluated. First, after replacing the original piezo-electric kinematic stage by a DC motor-based robot manipulator, experimentation showed that the speed and the precise motion control of pipettes were improved. Second, by introducing an X-Y stage into the system, to manipulate the Petri dish around the microscope's field of view, multiple microinjection speed was improved. Third, by using SSD template matching to track the injection pipette, rather than the cross-correlation template matching algorithm used in the original system, improvements were made to pipette localization. Under human control, this new semi-automated system gives improved microinjection performance metrics compared to previously obtained results. The system is also providing implicit human knowledge of the microinjection process via the human-control interface. It is the encoding of this knowledge that will lead to the first fully automated system. The semi-automated microinjection system is being tested and evaluated in the AMC at UNC-Chapel Hill.

We propose computational sensor networks as a methodology to exploit models of physical phenomena in order to better understand the structure of the sensor network. To do so, it is necessary to relate changes in the sensed variables (e.g., temperature) to the aspect of interest in the sensor network (e.g., sensor node position, sensor bias, etc.), and to develop a computational method for its solution. As an example, we describe the use of the heat equation to solve the sensor localization problem. Simulation and physical experiments are described.

The focus of this paper is the design and development of a semi-automated system for microinjection of embryonic stem cells into blastocysts. Semi-automation is achieved through treating cell microinjection as a computer game. In this first phase, cell manipulation and microinjection is carried out using a joystick and an interactive graphical user interface (GUI). For this system to be developed further, to achieve full automation, this first phase had to succeed. The interactive GUI records and displays: real-time video images of the cells; processed images of cells; and the microinjection strategies adopted by operators during cell manipulation through a real-time video processing algorithm. Experiments showed that this first phase of the research was successful. Future research develop knowledge-based controllers using machine-learning techniques as the research drives towards its goal of fully automated cell microinjection

This paper describes machine vision techniques that provide fast visual feedback for an automatic blastocyst microinjection system. The goals of the vision processing were to locate and track both the blastocysts and manipulation pipettes within the images and throughout the microinjection process. This was successfully accomplished using Hough transforms and cross-correlation template matching. Emphasis here is placed on a detailed description of the techniques applied to the algorithm to speed-up the matching and to enable real-time visual feedback. These techniques allowed the algorithm to be, on average, 857 times faster than the original algorithm. This converts into a 99.88% reduction in the template matching processing time

Transference of controllers evolved in simulation to real vehicles is an important issue in evolutionary robotics (ER). We have previously evolved autonomous navigation controllers for fixed wing UAV applications using multi-objective genetic programming (GP). Controllers were evolved to locate a radar source, navigate the UAV to the source efficiently using on-board sensor measurements, and circle around the emitter. We successfully tested an evolved UAV controller on a wheeled mobile robot. A passive sonar system on the robot was used in place of the radar sensor, and a speaker emitting a tone was used as the target in place of a radar. Using the evolved navigation controller, the mobile robot moved to the speaker and circled around it. The results from this experiment demonstrate that our evolved controllers are capable of transference to real vehicles. Future research will include testing the best evolved controllers by using them to fly real UAVs.

Smart Sensor Networks (S-nets) are groups of stationary agents (S-elements) which provide distributed sensing, computation, and communication in an environment. In order to integrate information from individual agents and to efficiently transmit this information to other agents, these devices must be able to create local groups (S-clusters). A leadership protocol that creates static clusters has been previously proposed. Here, we further develop this protocol to allow for dynamic cluster updating. This accommodates on-the-fly network re-organization in response to environmental disturbances or the gain or loss of S-elements. We outline an informal argument for the correctness of this revised protocol. We describe our embedded system implementation of the leadership protocol in simulation and using a colony of robots. Finally, we present results demonstrating both implementations.

This paper demonstrates the use of small area acoustic array technology as passive sonar for an autonomous mobile robot sound localization and direction control. Real-time target tracking is based solely on received audio signals.

This paper reports on research into miniature, direct drive, high force/torque motors to support insect-sized mobile robotic platforms. The primary focus is on scalable motors based on piezoelectric transducers. The contributions of this work include: (1) the design, analysis, and characterization of a miniature mode conversion rotary ultrasonic motor based on a piezoelectric stack transducer; this produced a static torque density of 0.37 Nm/kg, (2) a millimeter scale linear piezometer, constructed with a parallel arrangement of annular stressed unimorph piezoelectric transducers and passive latches, exhibited 0.23 N of blocked force, and (3) simulation data is presented that compares these motor concepts to commercial systems in the context of scalability. Results suggest that smaller versions of the rotary ultrasonic motor would possess a static torque density seven times that of a commercial 3-mm electromagnetic system. This technology shows promise for driving the platform.

We are engaged in the construction of a smart sensor network to monitor snow conditions and help determine avalanche probability in back country ski areas in the Wasatch mountains. These sensor networks will be comprised of devices having the ability to communicate, compute and sense the environment temperature, light intensity, pressure, and other properties. We have developed several distributed algorithms for such networks and demonstrated them in simulation and small-scale experiments. Our goal is to build and test a 50-100 node network through fall 2004, and to deploy it for experimental study this coming winter.

This paper describes the development and testing of a new evolutionary robotics research test bed. The test bed consists of a colony of small computationally powerful mobile robots that use evolved neural network controllers and vision based sensors to generate team game-playing behaviors. The vision based sensors function by converting video images into range and object color data. Large evolvable neural network controllers use these sensor data to control mobile robots. The networks require 150 individual input connections to accommodate the processed video sensor data. Using evolutionary computing methods, the neural network based controllers were evolved to play the competitive team game Capture the Flag with teams of mobile robots. Neural controllers were evolved in simulation and transferred to real robots for physical verification. Sensor signals in the simulated environment are formatted to duplicate the processed real video sensor values rather than the raw video images. Robot controllers receive sensor signals and send actuator commands of the same format, whether they are driving physical robots in a real environment or simulated robots agents in an artificial environment. Evolved neural controllers can be transferred directly to the real mobile robots for testing and evaluation. Experimental results generated with this new evolutionary robotics research test bed are presented.

This paper presents an overview of research on the production of nonwoven and tailored 3D structures for protective garments (such as those worn by fire fighters) using robotics and meltblown technology. In particular, the integration of robotics and a small-scale melt-blowing unit is discussed. This paper develops the framework and general motivation for the overall study and describes in detail the novel 3D-fiber application system developed using a seven-degree of freedom system. This system will be used with control algorithms developed at the NCRC to improve uniformity of the shaped fabric structure.

The production of protective garments is a complex process which generally involves nonwoven and tailored fabric structures. This paper presents an approach that integrates robotics with meltblown technology in addressing some of the problems associated with this garment production. In particular, a rule base 3D control is developed and an evaluation of the quality control of the 3D fabric structures formed using this system is performed using a basis-weight function analysis to measure uniformity. Melt-blowing of a polypropylene mixture was conducted by directly forming 3D fabrics on a mannequin mold. The results obtained show that when a constant-velocity control strategy is used during the fiber spraying phase, the basis-weight uniformity had a 16% coefficient of variation (CV). However, when a rule-based control strategy is applied using a displacement transducer for feedback and a set of fiber application rules for control, the CV is reduced to 11%. Thus the proposed system achieves the goal of improving uniformity of shaped fabric structures.

This paper addresses the field of knowledge-based systems, and in particular the sub-field of knowledge-based control systems. The rule-based approach used, particularly in its machine-learning or rule-induction mode, continues as a major theme in the emerging field of data mining - i.e. the extraction of usable insights from large databases.

This paper describes a qualitative modelling and simulation approach to designing and developing application independent control architectures for automated workcells. A qualitative modelling toolbox, called QMTOOL, was constructed using AI and object-oriented software to achieve this specification. QMTOOL models are built and simulated through interactive dialogue with the user. A model encapsulates the high-level knowledge of the human expert, and a simulation run can verify the design at the specification level. With QMTOOL, visual interaction, direct manipulation of objects, and symbolic computing mechanisms, are all available to the designer.>

The idea of competitive learning for pattern-recognition applications is introduced. A brief review of two competitive learning models, T. Kohonen's self-organizing feature maps (1982, 1989) and S. Grossberg's ART networks (1987), is presented. Neural-net-based partitioning algorithms for learning control are introduced. A simulation study, of these algorithms incorporated into the BOXES machine learning control system is reported. Simulation results are presented and performance comparisons are made, using the BOXES algorithm as the standard, with the new neural-net-based partitioning method. The original BOXES partitioning method of fixed threshold quantization of state-space variables was used in the BOXES algorithm learning trials.>