Fraunhofer Institute for Manufacturing Engineering and Automation

Predictions obtained by, e.g., artificial neural networks have a high accuracy but humans often perceive the models as black boxes. Insights about the decision making are mostly opaque for humans. Particularly understanding the decision making in highly sensitive areas such as healthcare or finance, is of paramount importance. The decision-making behind the black boxes requires it to be more transparent, accountable, and understandable for humans. This survey paper provides essential definitions, an overview of the different principles and methodologies of explainable Supervised Machine Learning (SML). We conduct a state-of-the-art survey that reviews past and recent explainable SML approaches and classifies them according to the introduced definitions. Finally, we illustrate principles by means of an explanatory case study and discuss important future directions.

Electrical impedance tomography (EIT) has undergone 30 years of development. Functional chest examinations with this technology are considered clinically relevant, especially for monitoring regional lung ventilation in mechanically ventilated patients and for regional pulmonary function testing in patients with chronic lung diseases. As EIT becomes an established medical technology, it requires consensus examination, nomenclature, data analysis and interpretation schemes. Such consensus is needed to compare, understand and reproduce study findings from and among different research groups, to enable large clinical trials and, ultimately, routine clinical use. Recommendations of how EIT findings can be applied to generate diagnoses and impact clinical decision-making and therapy planning are required. This consensus paper was prepared by an international working group, collaborating on the clinical promotion of EIT called TRanslational EIT developmeNt stuDy group. It addresses the stated needs by providing (1) a new classification of core processes involved in chest EIT examinations and data analysis, (2) focus on clinical applications with structured reviews and outlooks (separately for adult and neonatal/paediatric patients), (3) a structured framework to categorise and understand the relationships among analysis approaches and their clinical roles, (4) consensus, unified terminology with clinical user-friendly definitions and explanations, (5) a review of all major work in thoracic EIT and (6) recommendations for future development (193 pages of online supplements systematically linked with the chief sections of the main document). We expect this information to be useful for clinicians and researchers working with EIT, as well as for industry producers of this technology.

Modern agriculture is related to a revolution that occurred in a large group of technologies (e.g., informatics, sensors, navigation) within the last decades. In crop production systems, there are field operations that are quite labour-intensive either due to their complexity or because of the fact that they are connected to sensitive plants/edible product interaction, or because of the repetitiveness they require throughout a crop production cycle. These are the key factors for the development of agricultural robots. In this paper, a systematic review of the literature has been conducted on research and commercial agricultural robotics used in crop field operations. This study underlined that the most explored robotic systems were related to harvesting and weeding, while the less studied were the disease detection and seeding robots. The optimization and further development of agricultural robotics are vital, and should be evolved by producing faster processing algorithms, better communication between the robotic platforms and the implements, and advanced sensing systems.

Industrie 4.0 introduces decentralized, self-organizing and self-learning systems for production control. At the same time, new machine learning algorithms are getting increasingly powerful and solve real world problems. We apply Google DeepMind’s Deep Q Network (DQN) agent algorithm for Reinforcement Learning (RL) to production scheduling to achieve the Industrie 4.0 vision for production control. In an RL environment cooperative DQN agents, which utilize deep neural networks, are trained with user-defined objectives to optimize scheduling. We validate our system with a small factory simulation, which is modeling an abstracted frontend-of-line semiconductor production facility.

Abstract Purpose of Review This review provides a comprehensive overview of machine learning approaches for vision-based robotic grasping and manipulation. Current trends and developments as well as various criteria for categorization of approaches are provided. Recent Findings Model-free approaches are attractive due to their generalization capabilities to novel objects, but are mostly limited to top-down grasps and do not allow a precise object placement which can limit their applicability. In contrast, model-based methods allow a precise placement and aim for an automatic configuration without any human intervention to enable a fast and easy deployment. Summary Both approaches to robotic grasping and manipulation with and without object-specific knowledge are discussed. Due to the large amount of data required to train AI-based approaches, simulations are an attractive choice for robot learning. This article also gives an overview of techniques and achievements in transfers from simulations to the real world.

Stochastic model predictive control (MPC) strategies can provide guarantees of stability and constraint satisfaction, but their online computation can be formidable. This difficulty is avoided in the current technical note through the use of tubes of fixed cross section and variable scaling. A model describing the evolution of predicted tube scalings facilitates the computation of stochastic tubes; furthermore this procedure can be performed offline. The resulting MPC scheme has a low online computational load even for long prediction horizons, thus allowing for performance improvements. The efficacy of the approach is illustrated by numerical examples.

The transformation towards smart connected factories causes enormous changes in mechanical engineering industry starting from the development of cyber-physical production systems up to their application in production. Enterprise architectures already offer suitable methods to support the alignment of the internal IT landscape. New demands like customer involvement, iterative development and increased business-orientation arising with these digitized products require new approaches and methods. This paper presents the foundation and the first steps aiming at the development of a method for the holistic planning of the digital transformation in small and medium-sized mechanical engineering enterprises.

Thinking in terms of product life cycles is one of the challenges facing manufacturers today: Efforts to increase efficiency throughout the life cycle do not only mean extended responsibility of the parties concerned. Economically successful business areas can also be explored. Whether new service concepts are required, new regulations have been passed or consumer values are changing, the differences between business areas are disappearing. Life cycle management (LCM) considers the product life cycle as a whole and optimizes the interaction of product design, manufacturing and life cycle activities. The goal of this approach is to protect resources and maximize effectiveness by means of life cycle assessment, product data management, technical support and, last but not least, life cycle costing. This paper shows the existing approaches of LCM and discusses their prospects and further development.

Adopting digital transformation and Industrie 4.0 concepts is increasingly important for manufacturing companies acting in dynamic and competitive markets. In practice, however, these organizations struggle with implementing those concepts because Industrie 4.0 is rather a concept than a ready-to-implement solution. Additionally, its complexity hinders the achievement of successfully implementing Industrie 4.0 systems that truly incorporate all organizational aspects and levels. This paper introduces a framework to evaluate and guide Industrie 4.0 implementations, which build upon insights from capability maturity and alignment. Based on this, organizations can develop their own “roadmap” to increase the success of their Industrie 4.0 adoption.

International audience

The development of a mobile robot to assist people in their home is a long term goal of Fraunhofer IPA. As a vision of a future household product, the latest prototype, Care-O-bot <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">®</sup> 3, is equipped with the latest industrial state-of-the art hardware components and offers all modern multimedia and interaction equipment as well as most advanced sensors and control. Care-O-bot <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">®</sup> 3 has been presented to the public on several occasions where it distributed drinks to the visitors of trade fairs and events. Current developments aim at applying the robot in an eldercare facility in order to support the personnel in their daily tasks.

Sustainability improvements in industrial production are essential for tackling climate change and the resulting ecological crisis. In this context, resource efficiency can directly lead to significant advancements in the ecological performance of manufacturing companies. The application of Artificial Intelligence (AI) also plays an increasingly important role. However, the potential influence of AI applications on resource efficiency has not been investigated. Against this background, this article provides an overview of the current AI applications and how they affect resource efficiency. In line with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, this paper identifies, categorizes, and analyzes seventy papers with a focus on AI tasks, AI methods, business units, and their influence on resource efficiency. Only a minority of papers was found to address resource efficiency as an explicit objective. Subsequently, typical use cases of the identified AI applications are described with a focus on predictive maintenance, production planning, fault detection and predictive quality, as well as the increase in energy efficiency. In general, more research is needed that explicitly considers sustainability in the development and use phase of AI solutions, including Green AI. This paper contributes to research in this field by systematically examining papers and revealing research deficits. Additionally, practitioners are offered the first indications of AI applications increasing resource efficiency.

One of the major risk factors for global death and disability is alcohol, tobacco, and illicit drug use. While there is increasing knowledge with respect to individual factors promoting the initiation and maintenance of substance use disorders (SUDs), disease trajectories involved in losing and regaining control over drug intake (ReCoDe) are still not well described. Our newly formed German Collaborative Research Centre (CRC) on ReCoDe has an interdisciplinary approach funded by the German Research Foundation (DFG) with a 12-year perspective. The main goals of our research consortium are (i) to identify triggers and modifying factors that longitudinally modulate the trajectories of losing and regaining control over drug consumption in real life, (ii) to study underlying behavioral, cognitive, and neurobiological mechanisms, and (iii) to implicate mechanism-based interventions. These goals will be achieved by: (i) using mobile health (m-health) tools to longitudinally monitor the effects of triggers (drug cues, stressors, and priming doses) and modify factors (eg, age, gender, physical activity, and cognitive control) on drug consumption patterns in real-life conditions and in animal models of addiction; (ii) the identification and computational modeling of key mechanisms mediating the effects of such triggers and modifying factors on goal-directed, habitual, and compulsive aspects of behavior from human studies and animal models; and (iii) developing and testing interventions that specifically target the underlying mechanisms for regaining control over drug intake.

The proven ability of additive manufacturing (AM, also known as 3D printing) to fabricate complex components with substantial reductions in material wastage and reduced lead times has made it a key enabling technology for numerous important industrial applications. Of current AM processes, material jetting (MJ) is demonstrating considerable potential for producing multi-material, intricate, 3D components and systems with integrated functionality, with the additional benefit that the process can be easily integrated with other manufacturing procedures. However, material jetting of functional materials to produce advanced applications still poses numerous technological challenges which hinder its full industrial exploitation. These extend from the limited range of high-performance materials with consistent properties usable for MJ, through the need to enhance the process itself by optimizing the droplet formation process, tuning waveforms for specific processes, modification of substrate and substrate/jetted material interactions, the complexities of curing and post-processing procedures, and the challenge of characterizing the 3D printed parts. In this context, this article attempts to provide a comprehensive discussion of the principles and characteristics of the most recent material jetting technology and reviews the state-of-the-art research and development being conducted. This review identifies existing gaps with regards to high-performance UV-curable inks, printing behavior of non-Newtonian fluids, optimum jetting and curing strategies and effective measures for achieving high-precision MJ. Future work should bridge the aforementioned gaps in order to improve the performance of the technology, thereby making it more attractive for large-scale adoption by industry and increasing market acceptance and penetration of material jetting for advanced applications.

The division of floor plans or navigation maps into single rooms or similarly meaningful semantic units is central to numerous tasks in robotics such as topological mapping, semantic mapping, place categorization, human-robot-interaction, or automatized professional cleaning. Although many map partitioning algorithms have been proposed for various applications there is a lack of comparative studies on these different algorithms. This paper surveys the literature on room segmentation and provides four publicly available implementations of popular methods, which target the semantic mapping domain and are tuned to yield segmentations into complete rooms. In an attempt to provide new users of such technologies guidance in the choice of map segmentation algorithm, those methods are compared qualitatively and quantitatively using several criteria. The evaluation is based on a novel compilation of 20 challenging floor plans.

A novel innovative approach towards a marketable lab-on-chip system for point-of-care in vitro diagnostics is reported. In a consortium of seven Fraunhofer Institutes a lab-on-chip system called "Fraunhofer ivD-platform" has been established which opens up the possibility for an on-site analysis at low costs. The system features a high degree of modularity and integration. Modularity allows the adaption of common and established assay types of various formats. Integration lets the system move from the laboratory to the point-of-need. By making use of the microarray format the lab-on-chip system also addresses new trends in biomedicine. Research topics such as personalized medicine or companion diagnostics show that multiparameter analyses are an added value for diagnostics, therapy as well as therapy control. These goals are addressed with a low-cost and self-contained cartridge, since reagents, microfluidic actuators and various sensors are integrated within the cartridge. In combination with a fully automated instrumentation (read-out and processing unit) a diagnostic assay can be performed in about 15 min. Via a user-friendly interface the read-out unit itself performs the assay protocol, data acquisition and data analysis. So far, example assays for nucleic acids (detection of different pathogens) and protein markers (such as CRP and PSA) have been established using an electrochemical read-out based on redoxcycling or an optical read-out based on total internal reflectance fluorescence (TIRF). It could be shown that the assay performance within the cartridge is similar to that found for the same assay in a microtiter plate. Furthermore, recent developments are the integration of sample preparation and polymerase chain reaction (PCR) on-chip. Hence, the instrument is capable of providing heating-and-cooling cycles necessary for DNA-amplification. In addition to scientific aspects also the production of such a lab-on-chip system was part of the development since this heavily affects the success of a later market launch. In summary, the Fraunhofer ivD-platform covers the whole value chain ranging from microfluidics, material and polymer sciences, assay and sensor development to the production and assembly design. In this consortium the gap between diagnostic needs and available technologies can be closed.

In this contribution we present a robot assistant consisting of a mobile platform and a manipulator to support manual workplaces in production environments. We propose a general definition of robot assistants. Two applications for the robot assistant rob@work are described - the assembly of hydraulic pumps and the assistance of manual gas metal arc welding (GMAW). A description of the hardware and software architecture, the man-machine-interface, the task-planner and the path-planning of the robot assistant is given.

Purpose This review will describe the development of the Bionic Handling Assistant as well as the additive manufacturing (AM) process of robot grippers and its possibilities. Design/methodology/approach AM offers the chance to use the additive processes to produce highly flexible automation parts and systems as the Bionic Handling Assistant in small and medium quantities that can utilize a lot of design advantages provided by the process. Findings A lot of products of today and especially tomorrow could be produced by rapid manufacturing. New categories of products, such as the Bionic Handling Assistant, will occur. Originality/value In the paper, aspects of a visionary scenario for future productions are shown and demonstrated on the Bionic Handling Assistant.

There is a rapidly increasing interest in the use of synchrotron small-angle X-ray scattering (SAXS) for large-scale studies of biological macromolecules in solution, and this requires an adequate means of automating the experiment. A prototype has been developed of an automated sample changer for solution SAXS, where the solutions are kept in thermostatically controlled well plates allowing for operation with up to 192 samples. The measuring protocol involves controlled loading of protein solutions and matching buffers, followed by cleaning and drying of the cell between measurements. The system was installed and tested at the X33 beamline of the EMBL, at the storage ring DORIS-III (DESY, Hamburg), where it was used by over 50 external groups during 2007. At X33, a throughput of approximately 12 samples per hour, with a failure rate of sample loading of less than 0.5%, was observed. The feedback from users indicates that the ease of use and reliability of the user operation at the beamline were greatly improved compared with the manual filling mode. The changer is controlled by a client-server-based network protocol, locally and remotely. During the testing phase, the changer was operated in an attended mode to assess its reliability and convenience. Full integration with the beamline control software, allowing for automated data collection of all samples loaded into the machine with remote control from the user, is presently being implemented. The approach reported is not limited to synchrotron-based SAXS but can also be used on laboratory and neutron sources.

The present work aims to identify critical materials in water electrolysers with potential future supply constraints. The expected rise in demand for green hydrogen as well as the respective implications on material availability are assessed by conducting a case study for Germany. Furthermore, the recycling of end-of-life (EoL) electrolysers is evaluated concerning its potential in ensuring the sustainable supply of the considered materials. As critical materials bear the risk of raising production costs of electrolysers substantially, this article examines the readiness of this technology for industrialisation from a material perspective. Except for titanium, the indicators for each assessed material are scored with a moderate to high (platinum) or mostly high (iridium, scandium and yttrium) supply risk. Hence, the availability of these materials bears the risk of hampering the scale-up of electrolysis capacity. Although conventional recycling pathways for platinum, iridium and titanium already exist, secondary material from EoL electrolysers will not reduce the dependence on primary resources significantly within the period under consideration-from 2020 until 2050. Notably, the materials identified as critical are used in PEM and high temperature electrolysis, whereas materials in alkaline electrolysis are not exposed to significant supply risks.