Kira Barton

Paper Electronics Silicon micromodular transistors are integrated onto cellulose nanomaterial-coated paper with electrohydrodynamic jet-printed interconnects, establishing a manufacturing-relevant platform for high-performance paper-based microelectronics. More details can be found in the Research Article by Michael A. Filler and co-workers (DOI: 10.1002/admi.202501033).

ABSTRACT The integration of high‐performance microelectronics onto flexible, sustainable substrates is important for advancing eco‐friendly technologies. This study presents an approach for creating circuits comprised of silicon micromodular transistors on cellulose nanomaterial (CNM)‐coated paper substrates with silver nanoparticle interconnects formed using electrohydrodynamic jet (e‐jet) printing. The transistors exhibit on/off ratios of ∼10 7 , threshold voltages of 0.00 ± 0.06 V, subthreshold slopes of 102 ± 5 mV/decade, and peak effective mobilities of 430 ± 60 cm 2 V −1 s −1 . Cellulose nanofibril (CNF) and cellulose nanocrystal (CNC) coatings are evaluated for mechanical and electrical performance. CNF‐based coatings offer lower reverse saturation current despite higher surface roughness, while smoother CNC‐based coatings exhibit higher reverse saturation current. Wired micromodular transistors maintain performance under radii of curvature bending strain as small as 2.3 mm. A depletion‐load inverter serves as proof‐of‐concept micromodular circuitry. To the best of our knowledge, this work demonstrates a circuit assembly technique on paper that mitigates many of the trade‐offs of previous paper‐based platforms.

ABSTRACT High‐resolution printing offers promising avenues for packaging micro‐ and nanoscale modular electrical components, enabling hybrid, high‐performance circuits. The miniaturization of component interfaces imposes stringent requirements on printed interconnect resolution, conductivity, and structural robustness. This work systematically investigates the fabrication and characterization of submicron‐to‐micron scale (300 nm–3 ) metal nanoparticle interconnects, focusing on the interplay between printing parameters, multilayer deposition, and thermal sintering conditions. Silver (Ag) and gold (Au) colloidal inks are printed with controlled cross‐sectional geometries and sintered under dry air, forming gas, and nitrogen atmospheres. Correlations between geometry, sintering atmosphere, and interconnect resistivity reveal that oxygen‐rich and reducing environments promote nanoparticle coalescence, while resistivity and long‐term stability are strongly dependent on cross‐sectional area. Optimized processing yields Au interconnects with widths down to and resistivity of . Integration with micromodular transistor circuits demonstrates that interconnect geometry and device interface quality strongly influence series resistance, and analysis of failure modes identifies strategies to improve reliability. These results establish a framework for designing electrically robust, high‐resolution printed interconnects, enabling reliable integration into next‐generation microelectronic packaging and hybrid interconnection technologies.

Extrusion-based 3D concrete printing offers a transformative opportunity for waste-free fabrication of complex and materially optimized building components. However, the current reliance on manual observation and ad-hoc parameter tuning introduces significant variability into the process and limits the scalability of 3D concrete printing. Existing systems offer no predictive insight into failure events and remain dependent on trial-and-error methods, resulting in increased material consumption, higher costs, and prolonged production times. To address this, the paper introduces a comprehensive framework for process anomaly detection tailored to two-component extrusion-based 3D concrete printing systems. The proposed framework integrates systematic sensor instrumentation, heterogeneous data acquisition, and signal feature characterization to enable real-time monitoring of system health. Non-destructive sensors are deployed to measure environmental conditions, slurry pumpability, hydration behavior, and workability. Collected signals are analyzed using a suite of processing techniques — including frequency analysis, amplitude envelope detection, curve fitting, and magnitude deviation analysis — to extract critical features indicative of emerging anomalies. The methodology presented in this paper establishes a generalizable framework for anomaly detection, using data collected from physical experiments, where the specific metric values derived from the data are tied to the feedstock used in this study for validation. The paper results demonstrate the ability of the proposed methods to detect failure modes — such as pump clogging, material segregation, and two-component mixing head clogging — ahead of human operator detection. These findings confirm the viability of the proposed framework for anomaly detection.

Shared control, which merges the dynamic inputs of human drivers with vehicle automation, has attracted considerable attention due to its potential to enhance both safety and driver satisfaction. However, most existing shared control strategies are based on one-size-fits-all designs, neglecting the fact that the optimal level of sharing will typically depend on the individual driver and road characteristics. In light of these limitations, along with the observation that routes are commonly repeated, we propose an iterative learning control algorithm that divides a closed-loop driving circuit into discrete segments, enabling online estimation of the driver's performance over each segment. By adapting arbitration weights based on these driver-specific and segment-specific estimates, our method seeks to reduce lateral tracking errors, reduce path completion time, and increase secondary task performance. To validate this approach, we conducted driver-in-the-loop simulator tests with 20 participants, each driving repeatedly on the closed-loop circuit. The results demonstrate that our personalized strategy significantly improves driving performance.

To enable industrial robot autonomy without traditional manual programming, current approaches involve a carefully modeled environment or dedicated sensor feedback. This paper explores a novel alternative regime: on-demand sensing, in which a fleet of sensorless robots operating in un-modeled environments adapt to frequently changing repetitive tasks by requesting temporary access to a shared mobile sensor. A Hierarchical Sensor-Robot Control scheme is developed to enable an ad hoc team to cooperatively solve a task, at which point the sensor is dismissed while the robot repeats the task safely in open loop. An outer loop simultaneously optimizes the sensor pose and the parameters of an inner loop robot controller, which is encoded with potential fields. Simulation results demonstrate the algorithm converging for a realistic problem after just three outer-loop iterations.

Digital Twins (DTs) are key enablers of Smart Manufacturing, yet their adoption across the value chain is hindered by the lack of a standardized sharing framework. This paper addresses this challenge by identifying essential descriptive and qualitative elements of DTs based on standards and literature. Leveraging the Asset Administration Shell (AAS), it proposes a Submodel Template, which standardizes the packaging of DT models, interfaces, and computational and network requirements thus going beyond, and combining, existing AAS Submodels, i.e. for simulation models, to encapsulate the full multidimensionality of DTs. A case study on a Quality Monitoring DT (QM-DT) demonstrates the template’s ability to support seamless DT deployment, aggregation, and operation across heterogeneous manufacturing environments. Results show that the template enables structured transfer of subject matter expertise captured in DT models, real-time constraint support, and interoperability, laying the groundwork for improved DT integration and exchange.

This paper presents how game-theoretic level-k cognitive modeling, a framework for modeling the decision-making of agents with bounded rationality, can be used to describe the interactive policies within human-autonomy teams during collaborative tasks. The approach hypothesizes that prediction of human behavior can be used to enable autonomous systems to make better decisions. A case study is used to create a preliminary data library mapping human behavior to specific level-k strategies. Specifically, a participant completed laps on a shared control driving simulator while performing a secondary task. Between laps, the participant is informed of changes in the vehicle controller that impact the influence of the human input on the vehicle motion, reflecting changes in the belief the human holds about the autonomy, or changing their level-k. Using the library, the mapping between human behavior and level-k theory is evaluated through the classification of interactions between human-autonomy teams under directed scenarios.

Abstract Precision application within the coatings industry is a highly sought-after process with substantial gains in efficiency and performance. Its high degree of customizability allows for tailored solutions that meet specific requirements, further enhancing the overall quality and effectiveness of the coatings. Despite these advantages, precision application is a complex process influenced by numerous factors, and achieving the desired outcome is not always guaranteed. Typically, the printing performance is assessed after the process, with parameters being fine-tuned incrementally through trial and error. However, this routine of post-process evaluation can undermine its efficiency gains due to the repetitive cycles involved. This inefficiency highlights a clear gap in the industry for more advanced sensing and control systems that can construct an optimal input shape in real time. This work explores the experimental implementation of a control framework that synthesizes online model identification and model-based input shaping to achieve desired dimensions in uniform one- and two-dimensional printed patterns. Jetting frequency was selected as the primary input, allowing for the use of physics-based models to predict the process dynamics. Experimental validation of the proposed approach shows relatively consistent convergence of line width and film volume to reference values within 5 iterations. Additionally, minimal a priori information about the ink-substrate properties is required due to the incorporation of online model identification.

Abstract Extrusion-based printing is a widely used additive manufacturing (AM) process, typically controlled by motion inputs (stage speed) with a constant extrusion rate. However, relying solely on stage speed can introduce printing resolution errors, particularly in complex pattern regions like sharp angles. This paper develops a modeling framework for line width based on extrusion rate and stage speed, using the control volume concept. An optimal feedforward control framework is then designed to compute control inputs that accurately track line width while respecting system constraints. To refine these inputs, they are adjusted to match the system’s slowest time constant, yielding the final feed-forward signal, which significantly reduces errors in complex regions. To further address modeling limitations, an iterative learning control (ILC) framework is integrated with the feed-forward approach, mitigating errors across both printing and iteration directions. Experimental results strongly align with the model, demonstrating the framework’s reliability. The combined control system effectively enhances line width precision and pattern consistency across individual prints and repeated iterations, even in intricate designs.

This paper introduces a Multi-modal Diffusion model for Motion Prediction (MDMP) that integrates and synchronizes skeletal data and textual descriptions of actions to generate refined long-term motion predictions with quantifiable uncertainty. Existing methods for motion forecasting or motion generation rely solely on either prior motions or text prompts, facing limitations with precision or control, particularly over extended durations. The multi-modal nature of our approach enhances the contextual understanding of human motion, while our graph-based transformer framework effectively capture both spatial and temporal motion dynamics. As a result, our model consistently outperforms existing generative techniques in accurately predicting long-term motions. Additionally, by leveraging diffusion models' ability to capture different modes of prediction, we estimate uncertainty, significantly improving spatial awareness in human-robot interactions by incorporating zones of presence with varying confidence levels. Code: github.com/leob03/mdmp

Micromodular metal-HfOx-metal resistive random-access memories (RRAM) were fabricated, transferred to foreign substrates, and contacted using high-resolution electrohydrodynamic jet (e-jet) printing. The modular RRAM exhibited bipolar switching at low operating voltages and multi-level analog resistance states programmable using variable amplitude voltage pulses. These variable resistance RRAMs were also integrated with an amplifier and a bandpass filter to modulate their gain and corner frequencies. This work enables integration of high-quality RRAM devices in modular circuits and paves the way for microscale heterogeneous integration at the device level.

Supply chain (SC) risk management is influenced by both spatial and temporal attributes of different entities (suppliers, retailers, and customers). Each entity has given capacity and lead time to process and transport products to downstream entities. In disruptive events, lead times and capacities may vary, which affects the overall performance of SC. There have been many studies on SC disruption mitigation, but often without considering lead time and the magnitude of lateness. In this paper, we formulate a mixed integer programming (MIP) model to optimize SC operations via a routing and scheduling approach, to model the delivery time of products at different entities as they flow throughout the SC network. We minimize a weighted sum of multiple objectives that involve costs related to transportation, shortages, and delivery lateness. We further develop a Benders decomposition algorithm for speeding up the computation of the NP-hard MIP model. We also develop a discrete-event simulation framework to evaluate the performance of solutions to the MIP model under lead time uncertainty. Through extensive numerical studies, we show how the attributes of SC entities affect the performance, so that we can improve the SC design and operations under various uncertainties.

This letter introduces a novel probabilistic mapping algorithm, LatentBKI, which enables open-vocabulary mapping with quantifiable uncertainty. Traditionally, semantic mapping algorithms focus on a fixed set of semantic categories which limits their applicability for complex robotic tasks. Vision-Language (VL) models have recently emerged as a technique to jointly model language and visual features in a latent space, enabling semantic recognition beyond a predefined, fixed set of semantic classes. LatentBKI recurrently incorporates neural embeddings from VL models into a voxel map with quantifiable uncertainty, leveraging the spatial correlations of nearby observations through Bayesian Kernel Inference (BKI). LatentBKI is evaluated against similar explicit semantic mapping and VL mapping frameworks on the popular Matterport3D and Semantic KITTI data sets, demonstrating that LatentBKI maintains the probabilistic benefits of continuous mapping with the additional benefit of open-dictionary queries. Real-world experiments demonstrate applicability to challenging indoor environments.

The increasing complexity of modern manufacturing, coupled with demand fluctuation, supply chain uncertainties, and product customization, underscores the need for manufacturing systems that can flexibly update their configurations and swiftly adapt to disturbances. However, current research falls short in providing a holistic reconfigurable manufacturing framework that seamlessly monitors system disturbances, optimizes alternative line configurations based on machine capabilities, and automates simulation evaluation for swift adaptations. This paper presents a dynamic manufacturing line reconfiguration framework to handle disturbances that result in operation time changes. The framework incorporates a system process digital twin for monitoring disturbances and triggering reconfigurations, a capability-based ontology model capturing available agent and resource options, a configuration optimizer generating optimal line configurations, and a simulation generation program initializing simulation setups and evaluating line configurations at approximately 400x real-time speed. A case study of a battery production line has been conducted to evaluate the proposed framework. In two implemented disturbance scenarios, the framework successfully recovers system throughput with limited resources, preventing the 26% and 63% throughput drops that would have occurred without a reconfiguration plan. The reconfiguration optimizer efficiently finds optimal solutions, taking an average of 0.03 seconds to find a reconfiguration plan for a manufacturing line with 51 operations and 40 available agents across 8 agent types.

The capability of robots to rapidly adapt to new tasks without extensive reprogramming offers significant flexibility in reconfiguration of manufacturing processes to cope with unforeseen events. In modern manufacturing environments where numerous hardware and software systems exchange data with each other to perform a myriad of tasks, modularizing sub-systems and reusing commonly available information like product CAD models can increase robustness and efficiency of the reconfiguration. Yet, current approaches for robotic grasping tend to focus on standalone vision-based learning that often require either retraining to adapt to new object categories or massive dataset not available in manufacturing environments, making generalization challenging. This paper addresses the problem of exploiting available information, like CAD models, in manufacturing settings to efficiently generate a tractable set of grasps for known rigid objects, which can be directly applied to a wide class of robotic manipulations. In order to quickly produce diverse grasp configurations for arbitrary geometric models, we present GraspMixer, a combination of (1) an efficient offline sampler that utilizes specifications of a parallel-jaw gripper, and (2) a mapping function that fuses multiple features of a grasp to output a binary quality metric. During evaluation using physics-based simulations, a robotic gripper successfully executes 92.9% of all grasp configurations for 12 novel objects selected by GraspMixer. Among five different grasp sampling methods, GraspMixer also achieves the highest grasp success rate when performing table-top single object grasping under object pose uncertainty. The computation of this offline pipeline takes less than 1.0 minutes for each object without GPU hardware acceleration, which is comparable to or outperforms most of the benchmarks in the evaluation. Importantly, our framework exhibits impressive simulation-to-reality adaptation, achieving over 95% grasp success rate on previously unseen novel objects. All of these results are achieved with fewer than 10% of the samples typically used by other learning-based grasping techniques. Note to Practitioners—Modularization is a major theme in current manufacturing systems to increase efficiency. In this work, we introduce a new framework called GraspMixer, which is part of a larger manipulation and decision making architecture to enable versatile robotic manipulation in a manufacturing environment. The framework decomposes the task of reasoning about graspable local surfaces on object 3D models into sequentially connected sub-components. GraspMixer leverages information about objects and grippers, including their 3D models, materials, and inertial properties, which are available in a manufacturing environment. This enables our framework to automatically precompute grasping points on new objects that can be shared among multiple robots equipped with parallel-jaw grippers. GraspMixer synergizes with Internet of Things (IoT) and Cloud Computing platforms to efficiently scale up advanced robotic automation in manufacturing. Such a combination could provide greater flexibility in deploying advanced perception systems in a manufacturing environment to accelerate adaptation of the automation while saving computational resources of onboard processors within robots.

Micro-additive manufacturing describes a broad domain in 3D printing used to fabricate high-resolution patterns for printed electronics, biosensors, and labels. Material jetting, a process in which ink is printed and interacts with the surface in liquid form, is commonly used to make printed electronics. Despite advantages in material diversity and drop-on-demand capabilities, deviations in the volumes of the printed droplets can lead to poor device performance. Real-time feedback control is often infeasible due to fast jetting dynamics and high-resolution feature sizes. In this work, we consider iterative methods to address limitations in real-time monitoring and control actuation. Iterative model updating in the form of a piecewise linear approximation and nonlinear curve fitting is used to derive updated models of the process to enable feedforward parameter selections for subsequent patterns. A comparison with traditional model-based ILC is incorporated and various error metrics are reported. Among the standard and incrementally variable experimental conditions, the results indicate that the proposed approaches offer faster convergence and a significant reduction in transient error when transitioning between different widths for the reference line.

The determinants of human response to force perturbation include non-volitional mechanisms, such as biomechanics and reflex responses, as well as volitional means such as co-contraction and cognitive responses to haptic sensory feedback. Thus the modulation of neuromuscular admittance is receptive to instructions regarding these volitional strategies, even when the task remains constant. In this paper we investigate the influence of instruction on neuromuscular admittance when participants attempted to maintain the position of a manual control interface while experiencing unpredictable force perturbations. We found a clear distinction between trials where (N = 10) participants were instructed to stiffen their arm by co-contracting or respond to forces they felt by producing opposing forces. Our findings have implications on the design of shared control strategies between humans and automation.

Marine mammals rely on their flukes for propulsion. However, the forces generated by their foil-like flukes can not be measured directly due to the complexities of the marine environment. This study presents a data-driven modeling framework to investigate propulsive hydrodynamic forces during swimming. First, synthetic data was generated using a low-order simulation based on prior research to generate training data for model identification. The simulation models the two-dimensional translational motion (longitudinal and vertical) of the animal and approximates its fluking gait as a multi-linkage system. The propulsion force acting on the fluke is simulated using the principles of unsteady hydrodynamics and hydroelasticity. Subsequently, extended dynamic mode decomposition identifies a nonlinear model by lifting the original state-space into a higher-order nonlinear representation. The results demonstrate that the proposed method accurately estimates both the motion of the animal and the hydrodynamic forces exerted on it.

We present a framework to classify interactive states among bottlenose dolphins using data from a tagged individual and camera-based detections of others. A particle filter estimates continuous trajectories, while a heuristic rule based state classifier infers interactive states based on position, speed, and heading. Results reveal state-dependent movement differences and behavioral shifts over time. This approach enables fine-scale, data-driven analysis of social interactions in marine mammals.

Modern supply chain networks (SCN) are becoming increasingly complex, with vulnerable entities exposed to uncertain disruptions that affect local or global supply chain attributes. We model a stochastic mixed-integer program to minimize the overall cost of SCN design and operations, in response to lead-time and demand uncertainties following given probability distributions. We formulate a heterogeneous risk-aware model to trade off between cost and delay/shortage by considering different risk-attitudes amongst supply chain agents. In particular, we employ the Conditional Value-at-Risk (CVaR) as a coherent risk measure for quantifying risk while attaining solution tractability. We derive managerial insights from our numerical studies, finding the most benefit from diversifying agents in the root tier, since their disruptions affect all other tiers in the SCN. We find that as agents become more risk averse, the optimal solutions for key agents (such as assemblers), seek more backup suppliers and allocate extra capacities to achieve resiliency and reliability. Practitioners can use the outcomes of our framework and studies to guide SCN design considering heterogeneous risk attitudes between agents. Note to Practitioners—With growing uncertainties in global supply chains, inefficient responses to disruptions can lead to large penalties and long-term impacts such as customer dissatisfaction. This research is motivated by the challenges arising during the operations of supply chains under both lead-time and demand uncertainties. We employ optimization and centralized control approaches to optimize supply-chain network design as well as response strategies to disruptions, and our framework can handle heterogeneous risk preferences as it models the risk attitude of each individual entity or agent in supply chains. Our model can be utilized to completely or partially re-design resilient supply chains, to better prepare for unknown features and uncertainties. Our case study provides insights about risk-averse supply-chain designs that can reduce response cost, but increase initial investments on backups and redundancies.

This paper introduces a novel probabilistic mapping algorithm, LatentBKI, which enables open-vocabulary mapping with quantifiable uncertainty. Traditionally, semantic mapping algorithms focus on a fixed set of semantic categories which limits their applicability for complex robotic tasks. Vision-Language (VL) models have recently emerged as a technique to jointly model language and visual features in a latent space, enabling semantic recognition beyond a predefined, fixed set of semantic classes. LatentBKI recurrently incorporates neural embeddings from VL models into a voxel map with quantifiable uncertainty, leveraging the spatial correlations of nearby observations through Bayesian Kernel Inference (BKI). LatentBKI is evaluated against similar explicit semantic mapping and VL mapping frameworks on the popular Matterport3D and Semantic KITTI datasets, demonstrating that LatentBKI maintains the probabilistic benefits of continuous mapping with the additional benefit of open-dictionary queries. Real-world experiments demonstrate applicability to challenging indoor environments.

This paper introduces a Multi-modal Diffusion model for Motion Prediction (MDMP) that integrates and synchronizes skeletal data and textual descriptions of actions to generate refined long-term motion predictions with quantifiable uncertainty. Existing methods for motion forecasting or motion generation rely solely on either prior motions or text prompts, facing limitations with precision or control, particularly over extended durations. The multi-modal nature of our approach enhances the contextual understanding of human motion, while our graph-based transformer framework effectively capture both spatial and temporal motion dynamics. As a result, our model consistently outperforms existing generative techniques in accurately predicting long-term motions. Additionally, by leveraging diffusion models' ability to capture different modes of prediction, we estimate uncertainty, significantly improving spatial awareness in human-robot interactions by incorporating zones of presence with varying confidence levels for each body joint.

Micromodular n-channel metal-oxide-silicon transistors were fabricated, transferred to a foreign substrate, and adaptively interconnected using high-resolution electrohydrodynamic jet (e-jet) printed metal wires to create depletion-load nMOS inverters. The transferred transistors have effective electron mobilities approaching 500 cm<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$^{{2}} \cdot $ </tex-math></inline-formula> V<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$^{-{1}} \cdot $ </tex-math></inline-formula> s−1 and subthreshold swing as low as 82 mV/decade, while the nMOS inverters have gains close to 30. Detailed electrical characterization shows that e-jet printing does not impact transistor performance. Moreover, e-jet printing can accommodate variations in transistor placement, opening the door to systems that can correct manufacturing errors in real-time. This work sets the stage for on-demand microelectronics manufacturing with extreme customizability at the transistor level.

This paper explores the problem of automated robot program generation from limited historical data when neither accurate geometric environmental models nor online vision feedback are available. The Program Reuse Framework (PRF) is developed, which uses expert-defined motion classes, a novel data structure introduced in this work, to learn affordances, workspaces, and skills from historical data. Historical data comprise raw robot joint trajectories and descriptions of the robot task being completed. Given new tasks, motion classes are then used again to formulate an optimization problem capable of generating new open-loop, skill-based programs to complete the tasks. To cope with a lack of geometric models, a technique to learn safe workspaces from demonstrations is developed, allowing the risk of new programs to be estimated before execution. A new learnable motion primitive for redundant manipulators is introduced, called a redundancy dynamical movement primitive, which enables new end-effector goals to be reached while mimicking the whole-arm behavior of a demonstration. A mobile manipulator part transportation task is used throughout to illustrate each step of the framework.

Micro-additive manufacturing techniques pertaining to material jetting have demonstrated strong applicability in the fields of printed electronic and photonic devices. Nano-scale patterning has been achieved using electrohydrodynamic jet (e-jet) printing, which utilizes a series of high voltage pulses to precisely deposit ink in the form of printed patterns. Previous studies have successfully implemented learning control frameworks to achieve desired printing performance by adjusting the magnitude and timing of the voltage pulse. However, optimization of the input shape from a process-driven perspective has not been analyzed, and the risks of nozzle clogging and printing regime fluctuations remain as challenges in maintaining stable performance. Additionally, a knowledge gap persists in characterizing the effects of modulating the baseline voltage on temporal and volumetric dynamics of the jetting process, which serve to provide critical time constants in the design of the input shape. This work implements data-driven modeling techniques to quantify the effects of varying the pulsed and baseline voltages while developing a process-driven optimization algorithm for promoting stable jetting, furthering the foundation for implementation of control strategies for e-jet printing.

Humans are an essential part of Smart Manufacturing systems. While modeling machines based on real-time data and knowledge from subject matter experts has been widely explored, modeling humans has not. Modeling humans presents significant challenges due to the heterogeneity of individuals and the stochasticity that may be introduced through individual decisions based on events. We hypothesize that human behavior within a controlled environment and task will lie within a predictable scope that can give insight into future behavioral states of humans, such as fatigue. Human fatigue affects workers' productivity, safety, and performance in manufacturing systems. We define fatigue as measurable tiredness resulting from physical exertion. If we can predict when a human worker will become fatigued, breaks or task changes can be preemptively incorporated into system planning to maintain desired levels of productivity and safety. In this paper, we propose a modeling structure to describe fatigue in humans performing repetitive, manufacturing-like tasks. Using first order techniques, the model incorporates task context and physiological sensor data. A case study in which the participant completes a repetitive cable assembly task provides a demonstration of model development. We derived linear time-invariant state space models that are capable of predicting fatigue levels with high accuracy over a 20 minute time horizon.

Abstract Suction cups are commonly used to attach biologging tags to cetaceans, and interact mechanically with compliant integument, an organ primarily composed of skin and blubber. However, the impact of compliance on suction cup performance is difficult to predict because knowledge about in vivo integument mechanics is lacking. Here, an experimental approach is used to investigate the mechanical properties of common bottlenose dolphin ( Tursiops truncatus ) integument using a custom instrument, the static suction cup (SSCup), to collect data from both trained dolphins and wild individuals ( n = 17) during a static pose. Three loading profiles were applied at three sites to quantify nonlinear stiffness, hysteresis, and creep. The site at the dorsal fin insertion exhibited the highest stiffness, while sites posterior to the blowhole and above the pectoral fin showed greater energy dissipation during cyclic loading. Viscoelastic behavior was observed across all sites. Suction cup performance on a surrogate material with broadly similar compliance showed reduced performance compared to cups on rigid acrylic: the maximum applied force at detachment on acrylic (50 N) was twice as large as the compliant substrate (25 N). Site‐dependent compliance of integument may lead to varying performance of suction cups as an attachment method for tags.

Owing to the multi-factorial nature of fatigue, leveraging context to effectively monitor and intervene with fatigue symptoms presents a significant challenge. This paper aimed to understand how to incorporate context into designing mHealth systems for fatigue management. We conducted a two-week field study with 20 fatigue-vulnerable individuals using an activity-tracking sensor and self-reporting. We conducted data-prompted interviews to explore phenomena about participants’ fatigue experiences. Findings show a heterogeneous relationship between context and fatigue, which can be attributed to the phenomena that: (1) participants were influenced by multiple fatigue-inducing factors for different durations; (2) broad contexts moderated participants’ perceptions and coping strategies in response to local contexts; (3) the predictability and repetition of activities influenced participants’ fatigue perception and coping strategies. We propose a context-informed layer-based framework integrating these phenomena and discuss implications for designing fatigue management tools informed by our framework.

Speeds that minimize energetic cost during steady-state walking have been observed during lab-based investigations of walking biomechanics and energetics. However, in real-world scenarios, humans walk in a variety of contexts that can elicit different walking strategies, and may not always prioritize minimizing energetic cost. To investigate whether individuals tend to select energetically optimal speeds in real-world situations and how contextual factors influence gait, we conducted a study combining data from lab and real-world experiments. Walking kinematics and context were measured during daily life over a week (N=17) using wearable sensors and a mobile phone. To determine context, we utilized self-reported activity logs, GPS data and follow-up exit interviews. Additionally, we estimated energetic cost using respirometry over a range of gait speeds in the lab. Gross and net cost of transport were calculated for each participant, and were used to identify energetically optimal walking speed ranges for each participant. The proportion of real-world steady-state stride speeds within these ranges (gross and net) were identified for all data and for each context. We found that energetically optimal speeds predicted by gross cost of transport were more predictive of walking speeds used during daily life than speeds that would minimize net cost of transport. On average, 82.2% of all steady-state stride speeds were energetically optimal for gross cost of transport for all contexts and participants, while only 45.6% were energetically optimal for net cost of transport. These results suggest that while energetic cost is a factor considered by humans when selecting gait speed in daily life, it is not the sole determining factor. Context contributes to the observed variability in movement parameters both within and between individuals.

Robotic manipulation of deformable linear objects (DLOs) is an active area of research, though emerging applications, like automotive wire harness installation, introduce constraints that have not been considered in prior work. Confined workspaces and limited visibility complicate prior assumptions of multi-robot manipulation and direct measurement of DLO configuration (state). This work focuses on single-arm manipulation of stiff DLOs (StDLOs) connected to form a DLO network (DLON), for which the measurements (output) are the endpoint poses of the DLON, which are subject to unknown dynamics during manipulation. To demonstrate feasibility of output-based control without state estimation, direct input-output dynamics are shown to exist by training neural network models on simulated trajectories. Output dynamics are then approximated with polynomials and found to contain well-known rigid body dynamics terms. A composite model consisting of a rigid body model and an online data-driven residual is developed, which predicts output dynamics more accurately than either model alone, and without prior experience with the system. An adaptive model predictive controller is developed with the composite model for DLON manipulation, which completes DLON installation tasks, both in simulation and with a physical automotive wire harness.

Inside Front Cover In article number 2301407, Barton, Dasgupta, and co-workers introduced a subtractive patterning approach using electrohydrodynamic jet (e-jet) printing, where the printed ink results in material removal, rather than addition. The etching depth in the printed regions can be tuned with nanoscale precision. Using acid-based subtractive e-jet printing, sub-10 μm linewidths are patterned without the use of photolithography, which can expand the toolbox for nanomanufacturing of functional devices.

In the highly complex and stochastic global, supply chain environments, local enterprise agents seek distributed and dynamic strategies for agile responses to disruptions. Existing literature explores both centralized and distributed approaches, while most work neglects temporal dynamics and the heterogeneity of the risk management of individual agents. To address this gap, this paper presents a heterogeneous risk management mechanism to incorporate uncertainties and risk attitudes into agent communication and decision-making strategy. Hence, this approach empowers enterprises to handle disruptions in stochastic environments in a distributed way, and in particular in the context of multi-agent control and management. Through a simulated case study, we showcase the feasibility and effectiveness of the proposed approach under stochastic settings and how the decision of disruption responses changes when agents hold various risk attitudes.

Model inversion is a fundamental technique in feedforward control. Unstable inverse models present a challenge in that useful feedforward control trajectories cannot be generated by directly propagating them. Stable inversion is a process for generating useful trajectories from unstable inverses by handling their stable and unstable modes separately. Piecewise affine (PWA) systems are a popular framework for modeling complicated dynamics. The primary contributions of this article are closed-form inverse formulas for a general class of PWA models, and stable inversion methods for these models. Both contributions leverage closed-form model representations to prove sufficient conditions for solution existence and uniqueness, and to develop solution computation methods. The result is implementable feedforward control synthesis from PWA models with either stable or unstable inverses. In practice, feedforward control alone may yield substantial tracking errors due to mismatch between the known system model and the unknowable complete system physics. Iterative learning control (ILC) is a technique for achieving robustness to model error in feedforward control. To demonstrate the primary contributions' validity and utility, this article also integrates PWA stable inversion with ILC in simulations based on a physical printhead positioning system.

Efficient performance monitoring in production systems holds paramount importance as it enables organizations to optimize their manufacturing processes, enhance productivity, and maintain a competitive edge in the market. Typically, machine and system level performance monitoring systems are investigated independently, whereas an integrated approach that considers both levels can offer valuable insights and benefits. This paper introduces a data-driven approach for evaluating and improving the performance of production lines by monitoring the performance of both individual machines and their interactions as a system. The approach begins with a rigorous methodology for classifying machine states recorded by the Manufacturing Execution System (MES) into finer-grained substates, enabling a comprehensive analysis of machine cycle time variability. Subsequently, these substates are leveraged as a foundation for constructing performance monitoring models at both the machine and system levels, employing probabilistic automata for the machine level and logistic regression for the system level. The system-level performance monitoring model is constructed to predict a Flow metric that enables the prediction of abnormal behaviors and deviations from production targets. This data-driven approach serves as a foundational ingredient of a system-level digital twin, designed to provide production lines with insights that enable proactive implementation of measures aimed at optimizing overall manufacturing efficiency. Through an industrial test case from the automotive industry, the results demonstrate the capability of performance monitoring, capturing errors within confidence intervals, and establishing predictive cause-and-effect relationships between machines within the production system.

Abstract A customized atmospheric‐pressure spatial atomic layer deposition (AP‐SALD) system is designed and implemented, which enables mechatronic control of key process parameters, including gap size and parallel alignment. A showerhead depositor delivers precursors to the substrate while linear actuators and capacitance probe sensors actively maintain gap size and parallel alignment through multiple‐axis tilt and closed‐loop feedback control. Digital control of geometric process variables with active monitoring is facilitated with a custom software control package and user interface. AP‐SALD of TiO 2 is performed to validate self‐limiting deposition with the system. A novel multi‐axis printing methodology is introduced using x ‐ y position control to define a customized motion path, which enables an improvement in the thickness uniformity by reducing variations from 8% to 2%. In the future, this mechatronic system will enable experimental tuning of parameters that can inform multi‐physics modeling to gain a deeper understanding of AP‐SALD process tolerances, enabling new pathways for non‐traditional SALD processing that can push the technology towards large‐scale manufacturing.

Robotic manipulation of deformable linear objects (DLOs) is an active area of research, though emerging applications, like automotive wire harness installation, introduce constraints that have not been considered in prior work. Confined workspaces and limited visibility complicate prior assumptions of multi-robot manipulation and direct measurement of DLO configuration (state). This work focuses on single-arm manipulation of stiff DLOs (StDLOs) connected to form a DLO network (DLON), for which the measurements (output) are the endpoint poses of the DLON, which are subject to unknown dynamics during manipulation. To demonstrate feasibility of output-based control without state estimation, direct input-output dynamics are shown to exist by training neural network models on simulated trajectories. Output dynamics are then approximated with polynomials and found to contain well-known rigid body dynamics terms. A composite model consisting of a rigid body model and an online data-driven residual is developed, which predicts output dynamics more accurately than either model alone, and without prior experience with the system. An adaptive model predictive controller is developed with the composite model for DLON manipulation, which completes DLON installation tasks, both in simulation and with a physical automotive wire harness.

Real-world walking data offers rich insights into a person's mobility. Yet, daily life variations can alter these patterns, making the data challenging to interpret. As such, it is essential to integrate context for the extraction of meaningful information from real-world movement data. In this work, we leveraged the relationship between the characteristics of a walking bout and context to build a classification algorithm to distinguish between indoor and outdoor walks. We used data from 20 participants wearing an accelerometer on the thigh over a week. Their walking bouts were isolated and labeled using GPS and self-reporting data. We trained and validated two machine learning models, random forest and ensemble Support Vector Machine, using a leave-one-participant-out validation scheme on 15 subjects. The 5 remaining subjects were used as a testing set to choose a final model. The chosen model achieved an accuracy of 0.941, an F1-score of 0.963, and an AUROC of 0.931. This validated model was then used to label the walks from a different dataset with 15 participants wearing the same accelerometer. Finally, we characterized the differences between indoor and outdoor walks using the ensemble of the data. We found that participants walked significantly faster, longer, and more continuously when walking outdoors compared to indoors. These results demonstrate how movement data alone can be used to obtain accurate information on important contextual factors. These factors can then be leveraged to enhance our understanding and interpretation of real-world movement data, providing deeper insights into a person's health.