Search for: All records

Recognition of the role of extended defects on local phase transitions has led to the conceptualization of the defect phase, localized thermodynamically stable interfacial states that have since been applied in a myriad of material systems to realize significant enhancements in material properties. Here, we explore the kinetics of grain boundary confined amorphous defect phases, utilizing the high temperature and scanning rates afforded by ultrafast differential scanning calorimetry to apply targeted annealing/quenching treatments at high rates capable of capturing the kinetic behavior. Four Al-based nanocrystalline alloys, including two binary systems, Al–Ni and Al–Y, and two ternary systems, Al–Mg–Y and Al–Ni–Y, are selected to probe the materials design space (enthalpy of mixing, enthalpy of segregation, chemical complexity) for amorphous defect phase formation and stability, with correlative transmission electron microscopy applied to link phase evolution and grain stability to nanocalorimetry signatures. A series of targeted isothermal annealing heat treatments is utilized to construct a Time–Temperature-Transformation curve for the Al–Ni system, from which a critical cooling rate of 2400 °C/s was determined for the grain boundary confined disordered-to-ordered transition. Finally, a thermal profile consisting of 1000 repeated annealing sequences was created to quantify the recovery of the amorphous defect phase following sequential annealing treatments, with results indicating remarkable microstructural stability after annealing at temperatures above 90% of the melting temperature. This work contributes to a deeper understanding of grain boundary localized thermodynamics and kinetics, with potential implications for the design and optimization of advanced materials with enhanced stability and performance. 

Staphylococcus aureus is the leading cause of skin infections in the U.S., and its rapid evolution and resistance to antibiotics create a barrier to effective treatment. In this study, we engineered a composite membrane with bacterial cellulose and carbon nanotubes (BC-CNT) as an electroactive dressing to rapidly eradicate vancomycin-intermediate S. aureus. Nonpathogenic Komagataeibacter sucrofermentans produced the BC membrane at an air-liquid interface. Then, carboxyl-functionalized multi-walled CNTs were integrated into decellularized BC to create stable and electrically conductive BC-CNT dressings. The electric potential and ionic flux across BC-CNT were modeled and standardized via chronoamperometry for experimental validation. We found that treatment with electroactive BC-CNT increases S. aureus sensitivity to vancomycin and prevents macro-scale biofilm formation. The bactericidal efficacy of the composite membrane is consistent with electrochemical stress caused by voltage mediated with BC-CNT. After a single hour of combinatorial electrical and drug treatment, biofilm-forming capacity was inhibited by nearly 92 %. These results advance applications of electrochemistry in medicine and create a new direction to overcome S. aureus infections on skin and soft tissues. 

Given a swarm of limited-capability robots, we seek to automatically discover the set of possible emergent behaviors. Prior approaches to behavior discovery rely on human feedback or hand-crafted behavior metrics to represent and evolve behaviors and only discover behaviors in simulation, without testing or considering the deployment of these new behaviors on real robot swarms. In this work, we present Real2Sim2Real Behavior Discovery via Self-Supervised Representation Learning, which combines representation learning and novelty search to discover possible emergent behaviors automatically in simulation and enable direct controller transfer to real robots. First, we evaluate our method in simulation and show that our proposed self-supervised representation learning approach outperforms previous hand-crafted metrics by more accurately representing the space of possible emergent behaviors. Then, we address the reality gap by incorporating recent work in sim2real transfer for swarms into our lightweight simulator design, enabling direct robot deployment of all behaviors discovered in simulation on an open-source and low-cost robot platform. 

Automated behavioral measurement using machine learning is gaining ground in psychological research. Automated approaches have the potential to reduce the labor and time associated with manual behavioral coding, and to enhance measurement objectivity; yet their application in young infants remains limited. We asked whether automated measurement can accurately identify newborn mouth opening—a facial gesture involved in infants’ communication and expression—in videos of 29 newborns (age range 9-29 days, 55.2% female, 58.6% White, 51.7% Hispanic/Latino) during neonatal imitation testing. We employed a 3-dimensional cascade regression computer vision algorithm to automatically track and register newborn faces. The facial landmark coordinates of each frame were input into a Support Vector Machine (SVM) classifier, trained to recognize the presence and absence of mouth opening at the frame-level as identified by expert human coders. The SVM classifier was trained using leave-one-infant-out cross validation (training: N = 22 newborns, 95 videos, 354,468 frames), and the best classifier showed an average validation accuracy of 75%. The final SVM classifier was tested on different newborns from the training set (testing: N = 7 newborns, 29 videos, 118,615 frames) and demonstrated 76% overall accuracy in identifying mouth opening. An intraclass correlation coefficient of .81 among the SVM classifier and human experts indicated that the SVM classifier was, on a practical level, reliable with experts in quantifying newborns’ total rates of mouth opening across videos. Results highlight the potential of automated measurement approaches for objectively identifying the presence and absence of mouth opening in newborn infants. 

Autonomous surgical robots are a promising solution to the increasing demand for surgery amid a shortage of surgeons. Recent work has proposed learning-based approaches for the autonomous manipulation of soft tissue. However, due to variability in tissue geometries and stiffnesses, these methods do not always perform optimally, especially in out-of-distribution settings. We propose, develop, and test the first application of uncertainty quantification to learned surgical soft-tissue manipulation policies as an early identification system for task failures. We analyze two different methods of uncertainty quantification, deep ensembles and Monte Carlo dropout, and find that deep ensembles provide a stronger signal of future task success or failure. We validate our approach using the physical daVinci Research Kit (dVRK) surgical robot to perform physical soft-tissue manipulation. We show that we are able to successfully detect out-of-distribution states leading to task failure and request human intervention when necessary while still enabling autonomous manipulation when possible. Our learned tissue manipulation policy with uncertainty-based early failure detection achieves a zero-shot sim2real performance improvement of 47.5% over the prior state of the art in learned soft-tissue manipulation. We also show that our method generalizes well to new types of tissue as well as to a bimanual soft-tissue manipulation task. 

LLM chains enable complex tasks by decomposing work into a sequence of subtasks. Similarly, the more established techniques of crowdsourcing workflows decompose complex tasks into smaller tasks for human crowdworkers. Chains address LLM errors analogously to the way crowdsourcing workflows address human error. To characterize opportunities for LLM chaining, we survey 107 papers across the crowdsourcing and chaining literature to construct a design space for chain development. The design space covers a designer’sobjectivesand thetacticsused to build workflows. We then surfacestrategiesthat mediate how workflows use tactics to achieve objectives. To explore how techniques from crowdsourcing may apply to chaining, we adapt crowdsourcing workflows to implement LLM chains across three case studies: creating a taxonomy, shortening text, and writing a short story. From the design space and our case studies, we identify takeaways for effective chain design and raise implications for future research and development. 

Ensuring that AI systems do what we, as humans, actually want them to do, is one of the biggest open research challenges in AI alignment and safety. My research seeks to directly address this challenge by enabling AI systems to interact with humans to learn aligned and robust behaviors. The way in which robots and other AI systems behave is often the result of optimizing a reward function. However, manually designing good reward functions is highly challenging and error prone, even for domain experts. Consider trying to write down a reward function that describes good driving behavior or how you like your bed made in the morning. While reward functions for these tasks are difficult to manually specify, human feedback in the form of demonstrations or preferences are often much easier to obtain. However, human data is often difficult to interpret, due to ambiguity and noise. Thus, it is critical that AI systems take into account epistemic uncertainty over the human's true intent. My talk will give an overview of my lab's progress along the following fundamental research areas: (1) efficiently maintaining uncertainty over human intent, (2) directly optimizing behavior to be robust to uncertainty over human intent, and (3) actively querying for additional human input to reduce uncertainty over human intent. 

Tendon-driven continuum robot kinematic models are frequently computationally expensive, inaccurate due to unmodeled effects, or both. In particular, unmodeled effects produce uncertainties that arise during the robot’s operation that lead to variability in the resulting geometry. We propose a novel solution to these issues through the development of a Gaussian mixture kinematic model. We train a mixture density network to output a Gaussian mixture model representation of the robot geometry given the current tendon displacements. This model computes a probability distribution that is more representative of the true distribution of geometries at a given configuration than a model that outputs a single geometry, while also reducing the computation time. We demonstrate uses of this model through both a trajectory optimization method that explicitly reasons about the workspace uncertainty to minimize the probability of collision and an inverse kinematics method that maximizes the likelihood of occupying a desired geometry.