Offenburg University of Applied Sciences

This paper presents a dialog-based AI financial assistant designed to support small and medium-sized enterprises (SMEs) in interpreting their financial indicators. The system combines a modular multi-agent architecture with context engineering techniques (explicit agent roles, controlled function calls, and structured hand-offs) to analyze and explain financial data in natural language. A prototypical cash-flow analysis, embedded within a production platform, demonstrates how the assistant enables interactive communication of financial information, thereby improving access to managerial and financial knowledge for non-experts.

The open-circuit voltage curve (OCV curve) is a fundamental property of batteries. It is a signature of the battery chemistry and contains details on health and degradation modes. This article presents a novel method for operando diagnosis of the OCV curve using a voltage-controlled model (VCM). The method is based on a dynamic equivalent circuit model which uses measured voltage as input and simulates current as output. Theory and algorithm are developed for estimating the OCV curve from the difference between experimental and simulated current. The algorithm is demonstrated with four different lithium-ion battery cells that have different chemistries (metal-oxide and iron-phosphate positive electrodes), formats (pouch and prismatic), capacities (between 0.35 Ah and 180 Ah) and states of health (fresh and aged). Different operating protocols were used, spanning the range from 2-h single-cycle data up to 60-h multi-rate data and including dynamic and partial cycles. In all cases, the algorithm was able to successfully estimate the OCV curve. Mean absolute errors between estimated and true OCV curves were below 20 mV for each protocol and 10.1 mV in average. The algorithm thus allows to identify battery chemistries and to differentiate between fresh and aged batteries, without requiring low-current pseudo-OCV measurements. • New algorithm estimates the battery OCV curve using a voltage-controlled model. • Arbitrary protocols with dynamic voltage and current data possible. • Validated on four diverse lithium-ion cells (chemistry, format, capacity, age). • High accuracy achieved: Mean absolute error below 20 mV (average 10.1 mV). • Allows OCV estimation without needing slow, low-current pseudo-OCV measurements.

Eye-tracking-driven controls for assistive robotic arms provide people with severe physical disabilities with intuitive interaction opportunities.In this context, shared control can improve acceptance of the robot by partially automating task execution. Based on recent literature, we present a Wizard of Oz design for shared control driven by eye-tracking. This approach allows for the rapid exploration of user expectations to inform future design iterations. Two studies were conducted to assess user experience, identify design challenges, and determine ways to improve usability and accessibility. The first study involved people with severe motor disabilities (PSMD) and consisted of an online survey. The second study aimed to gain technical insights through direct robot interaction to provide a comprehensive overview of the findings.

Despite the substantial mass and moment of inertia of the trunk, its influence on minimizing metabolic cost in human locomotion is poorly understood. We show that self-selected trunk leaning in humans aligns with minimized metabolic costs during running. Utilizing an integrative experimental and musculoskeletal simulation approach, we reveal how trunk leaning redistributes joint moments and metabolic demands. We report estimates of muscle-specific changes in metabolic demands and contraction conditions for key muscles, including the hamstrings, uniarticular hip extensors and plantar flexors, reflecting the complex interplay between kinetic demands and muscle energetics. Our results show that leaning forward reduces knee extension moments but involves metabolic cost trade-offs. Our findings indicate that the observed U-shaped whole-body metabolic cost curve is attributed to intricate and offsetting changes in metabolic cost among multiple muscle groups across trunk leaning conditions, reflecting the fundamental impact of trunk kinematics on running biomechanics.

The following contribution presents a fully additively manufactured inductive displacement sensor applied to a model of a translationally movable robot flange. In addition to demonstrating the feasibility of 3D-printed coils, the focus is particularly on their inductive properties and the variation of inductance as a function of translational displacement. For this purpose, two coils are entirely additively manufactured and integrated into the flange of a 3D-printed robot model. Through translational movement of the flange, the overlap between the two opposing coils changes, enabling position detection by measuring the series inductance. The advantage of such additively manufactured approaches lies in their adaptability to diverse application requirements and their capability to realize complex geometries.

The human body consists of multiple joints with many degrees of freedom that open the context for biomechanical interventions in the form of recommendations and feedback, with the purpose of improving performance, reducing the risk of injuries, and so forth. These interventions sometimes involve loading trade-offs between joints or loading redistribution. Removing the load from one joint or segment (“native” joint) and transferring it to another joint (“step” joint) seem to need more consideration. Therefore, the aim of the current paper is to highlight the necessity of scientific communication from data collection to conclusions in the form of recommendations, presentations, and publications. The tendency to focus on specific joints and develop expertise hides the trade-off between joint loads, and sometimes, recommendations or feedback on the basis of biomechanical loading simply shift the risk of injury. To address this concern, several approaches are suggested, including improved marker sets, dataset availability, departmental diversity, research team collaboration, and the sharing of conferences and sessions.

This work presents a low frequency ground-penetrating radar (GPR) measurement system employing optimized Vivaldi antennas, targeting high penetration depth and enhanced range resolution. The radar integrates a modified edge-slot Vivaldi antenna operational from 75 MHz with a center frequency of 537.5 MHz and a bandwidth of 925 MHz. An initial GPR survey conducted in a ground-coupled mode over an asphalt surface successfully detects a buried utility pipe at approximately 2 m depth with good resolution. Detailed simulations, measurements and experimental results demonstrate the improved low frequency performance of the antenna, making it a suitable candidate for deep subsurface sensing.

Biochar is a carbon-rich material produced via pyrolysis that is increasingly recognized for its role in carbon sequestration, particularly through its application in agriculture and materials. However, accurately predicting the long-term persistence of biochar in the environment remains challenging. While incubation trials have been widely used to assess biochar degradation, their extrapolation beyond centennial timescales is uncertain. In this study, we evaluate the consistency between three physicochemical characterization methods that are considered as proxies for biochar persistence—hydropyrolysis (HyPy), solid-state electric conductivity (SEC), and elemental analysis to obtain molar hydrogen:carbon ratios. We produced 42 biochars from straw and wood using a continuously operated pilot-scale auger reactor at temperatures ranging from 400 to 800 °C under otherwise constant pyrolysis conditions. We then systematically analyzed the elemental composition, SEC and the fraction of biochar carbon that is resistant to HyPy (BCHyPy). Hydropyrolysis eliminates all free and covalently bound non-aromatic species and all aromatic species consisting of up to seven fused rings. Our results confirm that BCHyPy content increases with pyrolysis temperature and stabilizes above 600–680 °C, reaching >90% of total carbon in high-temperature biochars. Similarly, SEC increased exponentially with pyrolysis severity, correlating strongly with BCHyPy and H/C molar ratio. The latter has so far been used to predict biochar persistence. Our findings from a controlled temperature series of biochars highlight that SEC and BCHyPy could be useful proxies for parameterizing multi-pool decay models of biochars produced in practice.

Abstract Pancreatitis results from premature activation and impaired inactivation of pancreatic proteases, primarily trypsin, leading to self-digestion, tissue necrosis, fibrosis, and inflammation. SPINK1 is a pancreas-specific inhibitor of trypsin that prevents premature trypsin activation, and could be a candidate therapeutic. However, because of its small size, SPINK1 would be subject to rapid renal clearance, making it ineffective. To construct a long half-life therapeutic inhibitor of trypsin for pancreatitis treatment we fused this protein to the C-terminus of an IgG1 antibody Fc element, increasing the size to ~ 78 kDa, thereby exceeding the renal clearance threshold and providing for FcRn-mediated recycling out of cells. A non-glycosylated form of Fc-SPINK1 was expressed in the yeast Pichia pastoris. Fc-SPINK1 inhibits trypsin enzyme activity in vitro. The blood pharmacokinetics in mice are consistent with a three-compartment distribution model and a terminal half-life of ~ 3 days. In a cerulein-induced mouse model of pancreatitis, Fc-SPINK1 significantly ameliorated cell death and immune cell infiltration. We developed an automated image analysis technique to quantify pancreatitis-associated loss of tissue cohesion, and found that Fc-SPINK1 also reduced this effect. This study demonstrates the potential of Fc-SPINK1 as a rationally designed therapy for pancreatitis.