Search

We focus on enabling damage and tampering detection in logistics and tackle the problem of 3D shape reconstruction of potentially damaged parcels. As input we utilize single RGB images, which corresponds to use-cases where only simple handheld devices are available, e.g. for postmen during delivery or clients on delivery. We present a novel synthetic dataset, named Parcel3D, that is based on the Google Scanned Objects (GSO) dataset and consists of more than 13,000 images of parcels with full 3D annotations. The dataset contains intact, i.e. cuboid-shaped, parcels and damaged parcels, which were generated in simulations. We work towards detecting mishandling of parcels by presenting a novel architecture called CubeRefine R-CNN, which combines estimating a 3D bounding box with an iterative mesh refinement. We benchmark our approach on Parcel3D and an existing dataset of cuboid-shaped parcels in real-world scenarios. Our results show, that while training on Parcel3D enables transfer to the real world, enabling reliable deployment in real-world scenarios is still challenging. CubeRefine R-CNN yields competitive performance in terms of Mesh AP and is the only model that directly enables deformation assessment by 3D mesh comparison and tampering detection by comparing viewpoint invariant parcel side surface representations. Dataset and code are available at https://a-nau.github.io/parcel3d., Accepted at CVPR workshop on Vision-based InduStrial InspectiON (VISION) 2023, see https://vision-based-industrial-inspection.github.io/cvpr-2023/

Computer vision applications in transportation logistics and warehousing have a huge potential for process automation. We present a structured literature review on research in the field to help leverage this potential. The literature is categorized w.r.t. the application, i.e. the task it tackles and w.r.t. the computer vision techniques that are used. Regarding applications, we subdivide the literature in two areas: Monitoring, i.e. observing and retrieving relevant information from the environment, and manipulation, where approaches are used to analyze and interact with the environment. Additionally, we point out directions for future research and link to recent developments in computer vision that are suitable for application in logistics. Finally, we present an overview of existing datasets and industrial solutions. The results of our analysis are also available online at https://a-nau.github.io/cv-in-logistics.

Volume 2022, Issue 18, The trend towards heterogeneous, decentral systems in intralogistics results in the need for a concept to describe and virtualize assets to enable their interaction. The multi-layer concept of Cyber-Physical Intralogistics Systems (CPIS) is introduced. The system description (descriptive layer) defines the structure of the digital twins and the communication (virtual layer) of physical (robots, periphery) and logical assets (control systems, simulations). To implement this concept, an experimental environment was developed at the Institute for Material Handling and Logistics and the Karlsruhe Institute of Technology. It consists of physical components, such as models of mobile robots or manipulators, and further periphery, such as racks and charging stations. The environment is supplemented by simulations and control software. Use cases for CPIS are to be implemented and tested in this environment. Due to the easily accessible hardware components and the possible scaling of the systems in the simulation, implementation cycles can be reduced, and results can be achieved quickly without requiring a real-world intralogistics system. CPIS can be used to initialize an automated charging process or to exchange perceived position data of system participants. The primary goal is to enable a modular system, add new participants through plug-and-play, and make systems easily changeable., Der Trend zu heterogenen, dezentralen Systemen in der Intralogistik erfordert ein Konzept zur Beschreibung und Virtualisierung von Assets, um deren Interaktion zu ermöglichen. Das mehrschichtige Konzept der Cyber-Physical Intralogistics Systems wird eingeführt. Die Systembeschreibung (Descriptive Layer) definiert die Struktur der digitalen Zwillinge und der Kommunikation (Virtual Layer) von physischen (Roboter, Peripherie) und logischen Assets (Steuerungssysteme, Simulationen). Zur Umsetzung dieses Konzepts wurde am Institut für Fördertechnik und Logistiksysteme des Karlsruher Instituts für Technologie eine Versuchsumgebung entwickelt. Sie besteht aus physischen Komponenten, wie den Modellen mobiler Roboter oder Manipulatoren, und anderen intralogistischen Komponenten, wie Regalen und Ladestationen. Ergänzt wird die Umgebung durch Simulationen und Steuerungssoftware. In dieser Umgebung sollen Anwendungsfälle für CPIS implementiert und getestet werden. Durch die leicht zugänglichen Hardwarekomponenten und die mögliche Skalierung der Systeme in der Simulation können Implementierungszyklen verkürzt und Ergebnisse schnell erzielt werden, ohne dass eine reales Intralogistiksystem benötigt wird. CPIS können zur Initialisierung eines automatisierten Ladevorgangs oder zum Austausch von Positionsdaten aufgenommen durch Systemteilnehmer genutzt werden. Das primäre Ziel ist es, ein modulares System zu ermöglichen, neue Teilnehmer durch Plug-and-Play hinzuzufügen und Systeme wandelbar zu machen.

The data sets and regression models presented here are related to the article "Point and interval estimation of decomposition error in discrete-time open tandem queues". The data sets are the first to analyze the approximation quality of the discrete-time decomposition approach and contain independent and dependent (explanatory) variables for the analysis of decomposition error, which were obtained using discrete-time queueing models and discrete-event simulation. Independent variables are the utilization parameters of the queues, and variability parameters of the service and arrival processes. Dependent variables are decomposition error with respect to the expected value and 95-percentile of the waiting time distribution at the downstream queue. This article presents multiple linear regression and quantile regression to explain the variance of the dependent variables for tandem queues with equal traffic intensity at both queues and for tandem queues with downstream bottlenecks, respectively.

Volume 2020, Issue 12, Logistics and production systems are confronted with a highly volatile business environment, a situation which increasingly pushes common supply chain analytics approaches to their limits. Process mining is an emerging technique to provide insights into business processes as they are being executed. However, the application of process mining in cross-organizational context has not been conclusively researched. In a literature overview, we review a set of 34 papers on the application of process mining in supply chains and classify them according to a three-stage maturity model. We find the majority of academic publications (28 papers) to focus on the construction of cross-organizational process models, 5 publications to derive models for alerting deviations and recommending decision support, and 1 paper to focus on automatic adjustments of the system behavior. Based on these findings, we conclude that the exploitation of process mining will be a key competitive advantage in supply chain management in the upcoming years. This applies not only for the design and management of steady-state supply chains, but also for the rapid adaptation of new solutions in transient systems., Logistik- und Produktionssysteme sind mit einem äußerst volatilen Geschäftsumfeld konfrontiert, eine Situation, die die gängigen Ansätze der Supply Chain Analytics zunehmend an ihre Grenzen stoßen lässt. Process Mining ist eine neue Technik, um Einblicke in die Geschäftsprozesse zu gewinnen, während sie ausgeführt werden. Die Anwendung von Process Mining im organisationsübergreifenden Kontext ist jedoch noch nicht abschließend erforscht. In einem Literaturüberblick werden 34 Veröffentlichungen über die Anwendung von Process Mining im Supply Chain Management vorgestellt und nach einem dreistufigen Reifegradmodell klassifiziert. Die Mehrzahl der Publikationen (28 Veröffentlichungen) sind auf die Erstellung organisationsübergreifender Prozessmodelle konzentriert, 5 Veröffentlichungen stellen Modelle zur Warnung vor Abweichungen und zur Empfehlung von Entscheidungshilfen vor und eine Veröffentlichung beschreibt die automatische Anpassung des Systemverhaltens. Auf der Grundlage dieser Erkenntnisse wird geschlussfolgert, dass die Nutzung von Process Mining in den kommenden Jahren ein entscheidender Wettbewerbsvorteil im Supply Chain Management sein wird. Dies gilt nicht nur für die Gestaltung und das Management von stationären Lieferketten, sondern auch für die schnelle Anpassung neuer Lösungen in transienten Systemen.

Order picking systems are confronted with a volatile demand and short delivery time requirements. Manufacturing companies face the increasing variability requirements with Heijunka-levelling, one method of the Toyota Production System. The objectives of this publication are to develop a levelling concept for order picking systems, to analyse its performance based on a discrete-time analytical model and to develop a staffing algorithm determining the required workforce level in an order picking system with levelled order release. The levelling concept for order picking systems results from the existing models of Heijunka-levelling in the literature, which are adopted and expanded regarding the specific requirements of order picking systems. The order picking system with levelled order release is depicted as a discrete-time Markov chain. To analyse its performance, we derive several performance measures, such as service level, backlog duration and system utilisation, from the steady-state distribution of the Markov chain. The s taffing a lgorithm i s a binary search algorithm based on the Markov chain. The models developed in this publication enable a quantitative evaluation of the impact of several system parameters, such as variability of customer demand, workforce level and traffic intensity, on the performance measures of the order picking system. Furthermore, the staffing algorithm determines the workforce level which is required to guarantee a certain system performance, such as a service level of 99%, in an order picking system with levelled order release. By comparing levelled order release to FCFS-based order release strategies in a numerical example, we show the benefits of levelled order release.

Productive4.0 is a large-scale research project, which aims at using Industry 4.0 ideas and technologies to improve the design, manufacturing and distribution of high-tech products. Within this project, two work-packages provide models, which shall be used to design, optimize and control large supply chains, which consist of manufacturing and distribution elements. In our contribution, we give an overview about the current state of the usage and usability of stochastic, especially queueing models for the above mentioned purpose.

Volume 1 / 2018 27th Aachen Colloquium Automobile and Engine Technology 2018, Aachen, Germany, 8 Oct 2018 - 10 Oct 2018; Aachen : Aachener Kolloquium Fahrzeug- und Motorentechnik GbR 663-694 (2018)., Published by Aachener Kolloquium Fahrzeug- und Motorentechnik GbR, Aachen

Volume 2018, Issue 01, Development of safety-related control systems requires extensive testing to verify correct functionality of the developed system in all circumstances. In this work, a simulation model of a novel hoist concept is shown. The simulation model consists of a complete electromechanical model of the drivetrain, including the pole-changing induction motor, slipping clutch and the brake, coupled with a two-mass spring-damped model of the hoist, the load and the crane bridge. The simulation model enables model-assisted testing of a hoist supervision unit currently being developed as part of a research project. We discuss the taken approach to software testing using the developed model., Die Entwicklung von sicherheitsbezogenen Steuerungen erfordert umfangreiche Tests, um die korrekte Funktion des entwickelten Systems unter allen Umständen zu überprüfen. In dieser Arbeit wird ein Simulationsmodell eines neuartigen Hebezeugkonzepts gezeigt. Das Simulationsmodell besteht aus einem kompletten elektromechanischen Modell des Antriebsstrangs einschließlich polumschaltbarem Induktionsmotor, Rutschkupplung und Bremse, gekoppelt mit einem Zwei-Massen-Feder-Dämpfer-Modell des Hubwerks, der Last und der Kranbrücke. Das Simulationsmodell ermöglicht die modellgestützte Verifikation einer im Rahmen eines Forschungsprojektes entwickelten Hubwerksteuerung. Wir besprechen die Vorgehensweise beim Software-Testen anhand des entwickelten Modells.

Volume 2018, Issue 01, The Industrial Internet of Things (IIoT) promises to bring benefits for all parties, from the machine manufacturers and system integrators to the end users, through data analytics and intelligent machines enabled by increased connectivity. The connectivity, however, brings along the possibility of unauthorised intrusions, causing disturbances in production or intralogistic systems, or in the worst scenario, acting as unintended or malicious triggers for safety hazards. In this work, we present an approach for safety and security co-assurance for machinery. Current safety and security standardisation for industrial control systems and for machinery is discussed, as well as safety and security co-assessment methodology. On this basis, an updated machine lifecycle model and an approach for safety and security risk co-assessment and risk reduction is described. Initial results based on the approach are discussed for a mixed-criticality controller device currently under development., Das Industrial Internet of Things (IIoT) verspricht Vorteile für alle Beteiligten, von den Maschinenherstellern und Systemintegratoren bis hin zu den Endanwendern, durch Datenanalyse und intelligente Maschinen mit erhöhter Konnektivität. Die Konnektivität bringt jedoch die Möglichkeit des unbefugten Eindringens mit sich, wodurch Störungen in der Produktion oder in logistischen Systemen verursacht werden oder welches im schlimmsten Fall das unbeabsichtigte oder böswil-lige Auslösen von Sicherheitsrisiken bewirkt. In dieser Arbeit stellen wir einen Ansatz zur Sicherung von Maschinen, unter Berücksichtigung von Safety und Security, vor. Die aktuellen Sicherheitsnormen für industrielle Steuerungssysteme und für Maschinen werden ebenso erörtert wie die Methodik der Sicherheitsbeurteilung. Auf dieser Basis wird ein aktualisiertes Maschinenlebenszyklusmodell und ein Ansatz zur Bewertung und Reduzierung von Sicherheitsrisiken (Safety und Security) beschrieben. Erste Ergebnisse auf Basis des Ansatzes werden für eine in Entwicklung befindliche Mixed-Criticality-Steuerung diskutiert.

Volume 2018, Issue 01, Automated guided vehicles are designed for internal material transport in production and warehouse environments. To do this, transport orders must be assigned to the vehicles. In addition, the vehicles often have an electric drive. The batteries required for this are discharged during operation. Therefore, it must be decided when the vehicles must go to a charging station. This control option is often ignored and the vehicles are only sent for loading when the battery has (almost) completely discharged. In this work, a procedure that simultaneously solves the assignment of jobs and the decision when a vehicle should drive to a charging station is presented and evaluated. It is based on neural networks trained by genetic algorithms. The evaluation shows that the presented method delivers better results than a method that combines the "First-Come-First-Served" and the "Nearest-Vehicle-First" methods and in which the charging processes are controlled by a fixed battery threshold., Fahrerlose Transportsysteme dienen dem innerbetrieblichen Materialtransport im Produktion- und Lagerumfeld. Dafür müssen den Fahrzeugen Trans-portaufträge zugeordnet werden. Außerdem haben die Fahrzeuge oft einen elektrischen Antrieb. Die dafür nötigen Akkubatterien werden im Betrieb entladen, sodass zusätzlich entschieden werden muss, wann die Fahrzeuge zu einer Ladestation fahren sollen. Diese Steuerungsmöglichkeit wird oft ignoriert, sodass die Fahrzeuge nur zum Laden geschickt werden, wenn sich die Batterie (fast) vollständig entladen hat. In dieser Veröffentlichung wird ein Verfahren vorgestellt und evaluiert, das die Auftragszuordnung sowie die Entscheidung, wann ein Fahrzeug zu einer Ladestation fahren soll, gleichzeitig löst und auf von genetischen Algorithmen trainierten neuronalen Netzen basiert. Die Versuche zeigen, dass das vorgestellte Verfahren bessere Ergebnisse liefert als ein Verfahren, das eine Kombination aus „First Come First Served“- und dem „Nearest Vehicle First“-Verfahren darstellt und bei dem Aufladevorgänge nur eingeleitet werden, wenn die Fahrzeugbatterie einen Grenzwert unterschreitet.

Volume 2017, Issue 10, Automated guided vehicles are designed to autonomously transport material in production and warehouse environments. The loading/unloading process of the material on the vehicles occurs at dedicated stations, called material sources and destinations. Every time a vehicle is idle, a new transportation job, i.e. the transportation of some goods from a material source to a material destination, can be assigned to one of the vehicles, which represents the limiting resource. The policies, which are used for the job assignment, are several. In this paper, a new policy based on neural networks which were trained by genetic algorithms is proposed and evaluated. The results show that this new policy outperforms a policy which is a combination of the so called “First Come First Served” and the “Nearest Vehicle First” policy., Fahrerlose Transportsysteme werden häufig für den innerbetrieblichen Materialtransport im Produktions- und Lagerumfeld genutzt. Die Be- und Entladung mit Material findet an bestimmten Stationen, den Quellen und Senken, statt. Transportaufträge führen immer von einer Quelle zu einer Senke. Diese werden den Fahrzeugen, die die begrenzte Ressource im System darstellen, zugeordnet. Dafür gibt es unterschiedliche Verfahren. In dieser Veröffentlichung wird ein neues Verfahren vorgestellt und evaluiert, das auf von genetischen Algorithmen trainierten neuronalen Netzen basiert. Die Versuche zeigen, dass das vorgestellte Verfahren bessere Ergebnisse liefert als ein Verfahren, das eine Kombination aus „First Come First Served“- und dem „Nearest Vehicle First“-Verfahren darstellt.

In manufacturing systems, batch building processes are very common as goods are transported or processed in some steps in batches and have to be collected beforehand. In this paper we present a new method for the performance analysis of batch building processes in material flow systems with an underlying batch building rule. We first introduce the reader to a framework of model elements and associated computation methods which are well suited to model information flow and physical processes in manufacturing systems. With two real material handling case studies its applicability and benefits are shown. We then present the new model that extends avalable discrete queueing models by an element for the performance analysis of batch building processes under the capacitated timeout rule.

International audience; The purpose of this paper is to introduce a framework of model elements and associated computation methods which are well suited to model information flow and physical processes in manufacturing systems. The proposed elements, the so called stochastic finite elements, are based on a discrete representation of time, in order to render distribution assumptions unnecessary. In our research, we extend available discrete time queueing models by models for batch processes, which are very common in manufacturing systems, when goods are transported or processed in some steps in batches and in some other steps piecewise. We give a short review of these models.