DSpace@RPI (Rensselaer Polytechnic Institute)
Not a member yet
6809 research outputs found
Sort by
Large load manipulation and controller design with flexible joint manipulators
No full textMay 2022School of EngineeringRecent years have seen an increased interest in the use of flexible joint robots. This is due to wider use of collaborative and space robots. Flexibility is desired for collaborative robots because the inherent compliance makes interactions with humans safer. Space manipulator flexibility results from harmonic drives used to amplify the torque generated by the small actuators used to keep arm mass low for launch. Both situations require precise tracking of desired movements. However, this is difficult to achieve due to the inherent nonlinearity of flexible joint manipulators. There exist model-based and neural network based methods of controller design. However, these methods may result in complex controllers that require a high computational load. In space applications, there exist computational limitations that prevent the use of such control methods. For this reason, most space manipulators use simple controllers such as PID. Tuning the gains is difficult not only because of the arm nonlinearity but also because the arm operates both without and with large loads. Gain scheduling may be used to address the various operation modes. However, it would be desirable to have a single set of controller gains to prevent non-smooth transitions in controller output resulting from the gain changes. Whether designing one set or multiple set of gains, the tuning process can be very time consuming. Common methodologies require understanding of complex mathematics (such as for H-infinity methods) or a lot of active time from the control designer (such as hand-tuning). Therefore a methodology that can be implemented with basic controller design knowledge and consumes less of the controller designer's time is desirable. This thesis seeks to develop a methodology to generate robust, simple controllers for flexible joint robots that achieve the desired performance while operating both without and with a large load. The methodology consists of experiment design to gather experimental frequency response data, path generation, trajectory generation, and optimization problem formulation for controller design. The methodology is tested using a simulated arm and Rethink's Baxter robot. The simulated arm is an approximation to a space manipulator with seven degrees of freedom and joint flexibility. The simulation is performed using Simulink and Simscape Multibody. Baxter is a dual-arm industrial robot with each arm containing seven degrees of freedom. Baxter's joint flexibility results from series elastic actuators.Ph
Lipase Catalyzed Synthesis and Characterization of Poly (Glycerol-Sebacate)
No full textBiomacromolecules, 23, 398-408Note : if this item contains full text it may be a preprint, author manuscript, or a Gold OA copy that permits redistribution with a license such as CC BY. The final version is available through the publisher’s platform.This study demonstrated that immobilized Candida antarctica lipase B (N435) catalysis in bulk leads to higher molecular weight poly(glycerol sebacate), PGS, than self-catalyzed condensation polymerization. Since the glass-transition temperature, fragility, modulus, and strength for rubbery networks are inversely dependent on the concentration of chain ends, higher molecular weight PGS prepolymers will enable the preparation of cross-linked PGS matrices with unique mechanical properties. The evolution of molecular species during the prepolymerization step conducted at 120 °C for 24 h, prior to enzyme addition, revealed regular decreases in sebacic acid and glycerol-sebacate dimer with corresponding increases in oligomers with chain lengths from 3 to 7 units such that a homogeneous liquid substrate has resulted. At 67 h, for N435-catalyzed PGS synthesis, the carboxylic acid conversion reached 82% without formation of a gel fraction, and number-average molecular weight (Mn) and weight-average molecular weight (Mw) values reached 6000 and 59 400 g/mol, respectively. In contrast, self-catalyzed PGS condensation polymerizations required termination at 55 h to avoid gelation, reached 72% conversion, and Mn and Mw values of 2600 and 13 800 g/mol, respectively. We also report the extent that solvent fractionation can enrich PGS in higher molecular weight chains. The use of methanol as a nonsolvent increased Mn and Mw by 131.7 and 18.3%, respectively, and narrower dispersity (Đ) decreased by 47.7% relative to the nonfractionated product.https://login.libproxy.rpi.edu/login?url=https://doi.org/10.1021/acs.biomac.1c0135
Towards a Progression-Aware Autonomous Dialogue Agent
No full textRecent advances in large-scale language modeling and generation have enabled the creation of dialogue agents that exhibit human-like responses in a wide range of conversational scenarios spanning a diverse set of tasks, from general chit-chat to focused goal-oriented discourse. While these agents excel at generating high-quality responses that are relevant to prior context, they suffer from a lack of awareness of the overall direction in which the conversation is headed, and the likelihood of task success inherent therein. Thus, we propose a framework in which dialogue agents can evaluate the progression of a conversation toward or away from desired outcomes, and use this signal to inform planning for subsequent responses. Our framework is composed of three key elements: (1) the notion of a “global” dialogue state (GDS) space, (2) a task-specific progression function (PF) computed in terms of a conversation’s trajectory through this space, and (3) a planning mechanism based on dialogue rollouts by which an agent may use progression signals to select its next response
Lipase-catalyzed biodegradable polyol polyester synthesis and characterization
No full textAugust 2021School of ScienceThe lack of tough biodegradable elastomers underlies one of the most complicated and intriguing challenges in modern bioengineering, the search for affordable and effective mechanical substitutes for tissues. The efforts have included studies of bio-based polyol polyesters owing to their diversity in polymer compositions, structures, and corresponding physicochemical and biological properties. Over the past few decades, multiple polyol polyester architectures have been studied, however, while much progress has been made, relationships between monomer selection, the synthetic method used (including curing), and polyester properties remain unclear. Poly(glycerol sebacate), PGS, a polyol–polyester of glycerol with sebacic acid, has drawn significant attention given its advantageous properties such as ductility, biodegradability, and low biotoxicity. Current chemical synthesis methods rely on polycondensation polymerizations that are not selective, leading to gels unless reactions are quenched at low functional group conversions. The resulting low molecular weight macromers are amorphous and have a melting point below room temperature, which has limited applications. Another limitation of PGS in soft-tissue engineering is its rapid resorption time, which restricts its use to tissues that require relatively long-term mechanical support. In previous studies, lipase catalysis proved effective for the preparation of higher molecular weight glycerol-containing polyesters. In this work, catalysis by immobilized Candida antarctica Lipase B (Novozyme-435) of condensation polymerizations between a set of monomer compositions was studied to prepare PGS and its analogs with better physicochemical properties. It is expected that these developments, coupled with recent major progress in polyol polyester synthesis and ongoing fundamental research on the mechanism of biodegradation, should lead to the development of scalable, high-performance biomaterials that will undoubtedly revolutionize biomedical devices built for in vivo applications.
In this thesis, we address the aforementioned challenge by three aspects of effort: 1) lipase-catalyzed synthesis and characterization of poly (glycerol sebacate), 2) enzymatic polymerization of poly(glycerol-1,8-octanediol-sebacate) and its potential on electrospinning, and 3) effects of glycreol/1,8-octanediol molar ratio and curing conditions on the properties of poly(glycerol-1,8-octanediol-sebacate). This work seeks to explore the synthesis of non-crosslinked PGS of higher molecular weight relative to no-catalyst conditions via a regioselective lipase catalyst; the potential that a PGS analogue could be synthesized via N435 catalysis to possess properties that allow its electrospinning into nonwoven fiber scaffolds; and the fine-tuning of the physicochemical properties of poly(glycerol-1,8-octanediol-sebacate) via changing the molar ratio of glycreol/1,8-octanediol and curing conditions. With this effort, it will be possible to design new functional polyol polyesters of importance in biomedical applications where such structural variables are expected to better control mechanical properties, biocompatibility, degradation behavior.Ph
Exploring microbial growth of a model extremophile, archaeoglobus fulgidus, at elevated pressures
No full textMay 2019School of ScienceDeep-sea vent and subsurface microorganisms are metabolically diverse and often display unique adaptive strategies that operate under elevated pressure conditions. However, because high hydrostatic pressure (HHP) laboratory cultivation has not been widely adopted, knowledge of how these microorganisms function in native high-pressure environments is limited. To explore how elevated pressures affect the metabolism and physiology of deep-sea and subsurface microorganisms, growth of a model extremophile, Archaeoglobus fulgidus (type strain VC16), was investigated up to 98 MPa in batch cultures for both chemoorganoheterotrophic and chemolithoautotrophic metabolisms. A. fulgidus is an anaerobic, hyperthermophilic sulfate reducing archaeon, first isolated from a shallow marine vent but has been commonly identified in high-pressure marine environments (to 2-4 km below sea level, 20-40 MPa), including deep-sea hydrothermal vents, deep geothermal wells, and deep oil reservoirs. In heterotrophic HHP cultivation experiments, exponential growth was observed up to 60 MPa. Cell densities were comparable from 0.1-40 MPa, while lower cell densities were observed at 50 MPa and 60 MPa and growth was inhibited at 70 MPa. A. fulgidus displayed both piezotolerance and moderate piezophily under certain heterotrophic HHP conditions. In autotrophic HHP conditions, A. fulgidus displayed piezotolerance with similar growth rates and maximum cell densities observed at up to 40 MPa and little to no growth was observed at 60 MPa. A. fulgidus biofilm production was observed in certain heterotrophic conditions from 0.1-50 MPa under HHP batch cultivation conditions due to both low calcium concentrations in the growth medium and the presence of a stainless steel needle that created a nucleation site. This suggests that biofilm production here was a response to growth medium chemistry and surface area, and was not related to the elevated pressure conditions. Here, A. fulgidus was shown to grow, and in some cases also produce biofilm, over a range of elevated pressure conditions. To the extent of our knowledge, piezotolerance to HHP for both heterotrophic and autotrophic metabolisms have not been previously measured for a single species. A. fulgidus’ metabolic plasticity and capacity for biofilm production reflects adaptive mechanisms that lend insight into how this species thrives in extreme and fluctuating environments.Ph
Modeling efforts for improved meal prediction with application to blood glucose control
No full textAugust 2022School of EngineeringType 1 Diabetes Mellitus is a disease characterized by the loss of insulin production from beta cells in the pancreas, which results in unregulated blood glucose (BG). The condition is permanent, exacting a high toll on an individual in terms of both health outcomes and treatment burden. Acute risk of low BG includes coma, seizure, and death, while longer term risk of high BG includes damage to the circulatory and nervous systems. To mitigate the risk of both ends, individuals must constantly stay vigilant, regulating BG levels with insulin injections and constantly monitoring BG levels. The artificial pancreas aims to reduce both health and treatment burdens by automatic regulation of BG levels. It consists of an insulin pump that injects insulin, a continuous glucose monitor that provides BG measurements, and a control algorithm that calculates dosing decisions. The next step in the development of artificial pancreas systems is fully closed loop control around meals. Unannounced meals present a challenge for the control algorithm because of the uncertainty surrounding the presence and content of the meal in addition to the slow and irreversible action of insulin paired with the acute risk of low BG. We propose a meal model, developed on gold standard triple tracer data, that considers meal size and shape and explicitly estimates uncertainty within meals. To quantify the quality of prediction in the context of BG control, we propose a new metric that considers asymmetry of prediction error and assesses prediction distributions in addition to single point predictions. Using the proposed metric, we tune an extended version of the model on a large data-set of free-living patient data and compare to previous work. The proposed model is first compared against three simpler models using triple tracer meal data. Prediction root mean square error is improved by 11% relative to the next best non-linear model. A simple implementation of control for all models suggests improved control capability for the proposed model: the proposed model is the fastest to compensatefor a meal in 4 out of 6 cases and overcompensates the least (8% excess) in the worst case compared to other models (25% excess). The model is also validated on a large data-set by evaluating prediction capability and control performance. The proposed model improves predictions by 37% relative to previous work in terms of the proposed metric. In a retrospective simulation of control, the proposed model reduces clinical risk by 12% over previous work. An open source artificial pancreas system currently in use by many is Loop. Part of the presented work is an effort to introduce Loop into the control literature. We formulate the Loop control algorithm as a “coincidence point” model predictive control strategy paired with a linear state space model. We evaluate Loop in silico using error-prone scenarios and suggest improvements that can be made to the meal announcement functionality.Ph
Smart fracture fixation plate system for measuring callus stiffness in distal lateral femur fractures treated with fracture fixation plates
No full textDecember 2021School of EngineeringAbstract The purpose of this study was to adapt an off-the-shelf lateral femoral fracture plate into a smart fracture plate by implementing our smart fracture plate add-on that measures changes in callus stiffness.
The smart fracture plate add-on is comprised of a novel sensor that measures and transmits load wirelessly and a mechanical amplifier that attaches to the plate and converts the bending of the fracture plate into a transverse load that is then applied to the sensor. The bending of the plate is directly related to callus stiffness which is a biomechanical indicator of fracture healing progress. The sensor can be read wirelessly to determine changes in callus stiffness as an indication of healing progression. In this study, we used analytical and computational models of the fracture plate, mechanical amplifier, and sensor to design and optimize the smart fracture plate system design for improved sensitivity to callus stiffness changes. These models were the corroborated with in vitro experimental testing.
The mechanical amplifier design was optimized to maximize its sensitivity to changes in callus stiffness. The optimal combination of parameters was corroborated by an analytical model, computational model, and by experimental testing. The optimization of the force concentrator allowed the smart fracture plate system the ability to distinguish between phases of fracture healing under axial loads as low as 100 N or less than 15% body weight.
By measuring callus stiffness, the smart fracture plate add-on has the ability to distinguish between phases of healing and determine progression towards union or non-union. Early detection of progression towards non-union can also help prevent premature weight bearing which often causes catastrophic failure of the fixator. Current clinical methods of assessment are insufficient and other researched systems for measuring callus stiffness are not clinically viable. The smart fracture plate add-on is designed to be clinically viable by being able to adapt off-the-shelf fracture plates into “smart” plates.
Throughout the healing process, using the smart fracture plate add-on allows for the collection of unique clinical data to improve the patient’s course of care. Tracking and recognizing the trends of a patient’s healing progress early would allow the surgeon to adjust treatment and therapy and quickly and clearly see the effects in data collected at subsequent visits. The ability to objectively define healing progression has the potential to personalize post-operative care, optimize time to patient return to normal activity, prevent failures due to non-union, reduce healthcare costs, and promote better patient recovery and outcomes.M
Multimodal machine learning for human conversational behavior analysis
No full textAugust 2021School of EngineeringHuman behavior during interactive communication can be complex, consisting of an interplay of data streams in multiple modalities, from multiple parties occurring with different orders and timings. This thesis focuses on multimodal machine learning for interactive human behavior analysis. We develop computational algorithms to combine computer vision, natural language processing, and audio signal features to automatically interpret and reason about social communicative behaviors. Our methodology involves building deep learning architectures and temporal sequential models to understand the intra- and inter-modality dependencies of human behavior cues. The first part of this thesis involves predicting emergent leaders and dominant contributors in a group meeting scenario based on the frequency of certain action events, e.g. the percentage of time that a person is being looked at by other participants. This requires the estimation of visual focus of attention (VFOA) from frontal-facing videos, for which we developed a deep learning model to predict the visual target classes based on eye gaze and head pose. This model was also incorporated into an automatic meeting summarization algorithm to provide an importance score for the sentences to be extracted in the summary. To better model human interaction behavior in communication scenarios, the second part of the thesis addresses how to use multi-party co-occurrent visual events (e.g. how a person moves her body while being looked at by others) to predict Big-Five personality traits. For this objective, we correlated the frequency of co-occurrent visual events with each personality class to understand the importance of different visual features, and then applied a machine learning approach to predict the personality traits of participants in a group meeting. The third part of the thesis is aimed at further modelling temporal interdependencies across different modalities of features including visual, audio, and language. We developed an end-to-end multi-stream recurrent neural network (RNN) to help integrate non-synchronized features in a given time window. The algorithm is applied to predict participants’ social roles (e.g., Protagonist, Supporter, Neutral, Gatekeeper or Attacker) which can change frequently during the whole group meeting. Given long, untrimmed videos of human interaction, there are critical moments that we are particularly interested in, such as a short segment when the team reaches a milestone, or a time window when two participants have conflict with each other. The fourth part of the thesis aims at locating the boundary of these specific moments from the original video. For this objective, we constructed a multimodal moment localization framework that takes a natural language query and the video content as input, and outputs the critical moment that semantically matches the query. We designed a temporal convolution module to explore the relationship between meaning in the query and the interaction between neighboring video frames. Moving beyond the estimation of specific social signals such Big-Five personality or social emotional roles, we consider a more data-driven approach in which we develop a multimodal deep learning model to automatically reason about human interactive behavior in a natural way. Specifically, instead of classifying manually defined categories of one single social signal, in this part of our work, we focus on visual question answering, e.g., automatically answering questions like “How is the man who is not being blamed responding to the situation?” from a multiple choice list. We developed a temporal attention-based model to highlight the critical moment in the video content and the keywords in the question sentence for better abstracting the important information in the given multimodal materials, and we capture the cross-modal dependencies using a consistency measuring module.Ph
Rendering snow : a light transport model for compressed anisotropic granular media
No full textAugust 2021School of ScienceSnow is a complex material that can take on many different visual properties based on the structure and shape of the individual ice grains that compose it. Individual ice grains dictate the surface texture and the transport of light through the medium as an aggregate, but simulating at this detail is computationally expensive. Existing methods for rendering granular media address this cost by approximating light transport through randomly oriented grains and only use specific individual grains to resolve surface-level details. However, they assume that grains are approximated well as a packing of non-overlapping bounding spheres, which is not always the case for snow (i.e. the classic snowflake shape). I present a light transport model that relaxes the non-overlapping requirement and enables compressed packings for highly non-spherical, anisotropic, grain shapes. It can produce snow objects that are progressively denser in appearance, both in surface detail and light transport, from sparse packings to highly compressed packings across several grain types. This model and the geometric basis I establish for compressed packings is also promising for extending the state-of-the-art granular media framework to support compressed packings of anisotropic grains.M
Low temperature and impurity compensated gallium antimonide crystal growth for nonlinear optical applications
No full textMay 2022School of EngineeringGaSb is a III-V compound semiconductor substrate material most suitable for the epitaxial growth of antimonide based quantum well (QW) and super-lattice (SL) layer photodetector focal plane array (FPA) structures due to its close lattice parameter with the device layers. For back side illumination devices, the substrate material must have high optical transmission. Unfortunately, GaSb has unusual optical transmission characteristics due to the presence of high density of equilibrium point defects (native defects such as vacancies and antisites) which makes the substrates practically opaque to radiation of below bandgap wavelengths where the devices operate. The GaSb substrate must be thinned after device fabrication complicating the device packaging process and often resulting in the introduction of defects in the device layers. Optically transparent GaSb is highly desirable for these applications as well as other infrared optical technologies. GaSb also has excellent nonlinear optical properties that makes it attractive for optical power limiters (OPLs). When integrated with the FPAs, it could provide protection to the QW and SL sensors from threats posed by laser based weapons. For OPLs, optically transparent, thick (mm-cm scale) GaSb substrate is necessary. In this research, two fundamental crystal growth approaches have been experimented for enhancing the optical transmission of GaSb. In the first method, crystal growth of GaSb from liquid phase at low temperature has been conducted from gallium rich non-stoichiometric solution with growth temperature in the range of 400-600 oC. Lowering growth temperature is expected to reduce the concentration of equilibrium point defects (vacancies and antisites), thus enhancing the optical transmission of GaSb. The effects of temperature gradient, synthesis duration and solution cooling rate on the optical transmission of GaSb has been studied. To eliminate the incorporation of gallium inclusions in the crystals, a continuous solute feeding process has been developed. Using the solute feeding process, 2-3 mm thick crystals of 20 mm diameter GaSb have been successfully grown at temperature as low as 500 oC, which is approximately 200 oC lower than the melt growth temperature of GaSb with growth rates in the range of 0.2-0.5 mm per hour. However, the growth rate used was found to be high and the resultant GaSb wafers exhibited optical scattering from gallium and antimony based metallic inclusions frozen in the bulk matrix. Challenges in GaSb growth at such low temperatures will be discussed along with the mitigating strategies. In the second method, impurity doping with n-type dopant tellurium (Te) and p-type dopant zinc (Zn) was experimented for the first time to compensate for the p-type native defects in GaSb crystals. Past research on n-type doping using Te had limitations. The new co-doping approach using Te and Zn provides the flexibility to alter the Fermi level position in GaSb and hence the optical transmission can be enhanced by avoiding the ionization of the native defect levels (vacancies and antisites).
Undoped GaSb exhibit relatively high below bandgap single photon absorption coefficient, α ≈ 10-20 cm-1 at the wavelengths of interest for OPLs (1.9 - 3 μm), primarily due to high concentrations (~ 1017 cm-3) of electrically active native defects due to Ga antisite (GaSb) and Ga vacancies (VGa). By optimizing the Zn and Te concentrations respectively to 6 x 1017 cm-3 (Zn) and 3 x 1017 cm-3 (Te), this research has demonstrated a significant reduction in below band gap optical absorption coefficient by 8-10 cm-1. This is a 50-80% reduction in the absorption coefficient across the spectra studied. The Zn and Te exhibit shallow energy levels compared to the native defect levels in bandgap. In addition to the reduction of optical absorption, free carrier concentration level also decreased by a factor of 15-20 compared to undoped GaSb. This effectiveness of the co-doping approach for the enhancement of optical and electrical properties of GaSb has been attributed to the lowering of the Fermi level by 21.9 meV from the undoped GaSb Fermi level. The shift in Fermi level has been theoretically postulated as a result of the Zn and Te atoms occupying a significant portion (30-50%) of the GaSb and VGa defects. In addition, the lowering of the Fermi level relative to the valence band resulted in the incomplete ionization of the GaSb and VGa defect energy levels. This novel co-doping approach can be further exploited to engineer the Fermi level and equilibrium defect concentration independently to create optically transparent GaSb and other semiconductor compounds.Ph