Labor Parallele und Verteilte Systeme
Refine
Year of publication
Document Type
- conference proceeding (article) (21)
- Part of a Book (7)
- Article (5)
- Book (1)
Is part of the Bibliography
- no (34)
Keywords
- Higher-Order Components (3)
- Artificial intelligence (2)
- Buildings (2)
- Concurrent computing (2)
- Conferences (2)
- CoreGRID Component Model (GCM) & Fractal (2)
- Java (2)
- Middleware (2)
- Modeling (2)
- NAVIGATION (2)
Institute
Begutachtungsstatus
- peer-reviewed (13)
Recent advances in robot middleware, such as the Robot Middleware Framework (RMF) in the Robot Operating System 2 (ROS 2), support robot applications with an increasing complexity: Delivery tasks can, e. g., be accomplished in multi-level buildings, since the RMF includes a lift adapter for communicating with a variety of elevators. However, the specifications of the actual building levels often adhere to less flexible formats. In many cases serialized data are used for describing a workplace and these data are included in the Universal Robot Description Format (URDF) specification of the robot or in a separate (e.g., YAMLformatted) configuration. This information is raw to the extent that the content is limited to coordinate lists. Semantics, such as the purpose of a room, the relation between rooms and tasks or priorities are either not included at all or the specification is done in an undocumented way. Although a semantic understanding is primarily necessary in many robot applications, ROS 2 location descriptions demand the manual (i. e., text-based) insertion of the required functional information, while tools focus on specifying physical shapes and sizes only. Building Information Modeling (BIM) tools, in contrast, as they are used for the construction and the maintenance of modern buildings, provide the necessary means for combining spatial information with semantic annotations in a structured manner. Therefore, we propose to establish a toolsupported translation of BIM data to the robotics domain. For illustration purposes, we show how a knowledge tree, with all the required positioning data for a robot task, can be automatically derived from an annotated building specification.
Movement sensors are used, e.g., for automated lighting or for HVAC automation. Depending on the type of sensor (ultrasonic, radar, etc.), not only the presence but also the number of people in a building can be determined. ML models trained with movement data can support the interpretation of local movement patterns: Situated intelligence means putting elementary data, such as step sensor measurements, together and learning bottom-up which movement patterns may indicate processes such as crowd formation or hurried motion. Such dynamic inference mechanisms are useful in guidance systems for key public venues, including shopping malls or tourist attractions. Compared to video-based automation, situated AI is more adaptable and can also be used where filming is not allowed due to privacy. In this work, we discuss privacy-aware tracking for environments using floor sensors and training data we built synthetically, but realistically, by imitating human musculoskeletal biomechanics. For two feasibility case studies, one in a grocery store and one on a city bus, we explain how we track unidentified people and derive limited behavioral indicators only from foot movements in a narrow area that we covered with a sensor map.
Despite the versatility of today’s LLMs, maintaining control over automated systems via textual instructions remains challenging due to the ambiguity of human language. When we restrict our attention to automated guided vehicles (AGVs), i. e., mobile robots, the difficulty of translating commands like "come over" can be simplified to specifying the "over" precisely enough that the machine can follow. For specifying motion targets, the Industry Foundation Classes (IFC), used in architecture, or a Universal Scene Description (USD), as popular in visual effects (VFX) design, are definitely more accurate than prompting an AI, particularly concerning indoor locations. However, these data formats represent descriptive languages and are therefore not widespread in modern robotics, where most applications are built upon the robot operating system (ROS 2), which rather follows an instructive communication paradigm. Thus, a translation of floor-, room-, door- and stair-descriptions into a stream of tokens, holding primarily directions, is desirable. Whilst parsing that stream, geometric properties can be mapped to a spatial model. The Cognitive Robot Abstract Machine (CRAM) notation, extending the functional programming language Lisp by macros for controlling robots, provides a suitable syntax for describing such models as motion plans. Therefore, we present a converter from IFC to CRAM which we have built using the Unix tools Flex and Bison. For an example IFC file, we also explain the steps necessary for letting a ROS 2-compliant robot autonomously follow the CRAM motion plan resulting from such a conversion.
Die serienmäßige Produktion von Elektrolysezellen befindet sich im Anfangsstadium, sodass belastbare Erfahrungen für standardisierte Produktionssysteme fehlen. Die Planung ist daher durch hohe Unsicherheit geprägt. Unterschiedliche Gestaltungsmöglichkeiten der Produktionssysteme müssen frühzeitig bewertet und ihre Leistungsfähigkeit mit dynamischen Methoden, wie der Simulation, abgesichert werden. Anhand eines von den Autoren entwickelten Simulationsframeworks mit automatisierter Modellgenerierung auf Basis der Petri-Netz-Logik zeigt dieser Beitrag die Möglichkeit der Simulation von Produktionssystemen in der Planung. Ein besonderes Leistungsmerkmal des Frameworks ist, dass zunächst unübersichtlich generierte Modelle mit einem Graphenalgorithmus so angeordnet werden, dass ihre Funktionsweise aus der Darstellung nachvollziehbar wird.
Due to the recent surge in bear attacks, affected municipalities have taken various measures to drive out the intruders. Political considerations include, in many places, the expansion of shooting permits. More and more regional governmental authorities invest in technologies for efficient bear hunting, such as IP cameras and computer vision. From an ecological perspective, such procedures, which are all solely aimed at decreasing the bear population, are questionable, since ecosystems can benefit from a peaceful coexistence of bears with humans: Bears remove animal carcasses and thus return vital nutrients to the soil and prevent the spread of diseases. Moreover, the bears keep the deer population in check, which is advantageous for the forest vegetation.
Conflicts start, once the bears intrude populated areas, which are often rural villages and suburbs with a low population density. Naturally, bears do not see humans as prey and respect their size. When the animal attacks, this happens rather for territorial or protective reasons, e. g., after an unexpected encounter with humans unsettled it. In this work, we analyze means to eradicate the danger: We dynamically apply Weiszfeld’s algorithm for finding the shifting geometric median between time-windowed black bear sightings and use the results for implementing an early warning system. In addition, we weigh the spotted locations and predict the expected migration behavior between bear habitats using a simulation with intelligent agents. As a case study, we apply our software to data from Romania and Fukushima and identify the positions where bear repelling facilities should be placed most effectively.
Smart retail technologies save grocery store operators a lot of work. At the same time, these technologies produce valuable data for building sustainable and economical inventory management strategies. AI models can be trained for sales forecasting using the data. The forecasts support the provisioning of fresh food over the whole week and help reducing food waste. In this paper, we present a Web portal which we developed to allow grocery store operators experiments with AI models revealing interrelations between observed and anticipated customer behavior. Clickable diagrams facilitate the exploration of data sets combining historical data and synthetically generated data. Pricing and ordering can be adapted accordingly to the simulated forecasts. By means of a case study, we show that our simulations are not only useful for predicting future sales but for other smart retail tasks as well.
Angesichts eines dynamischen industriellen Umfelds müssen produzierende Unternehmen flexibel bleiben und sich kontinuierlich anpassen, um ihre Wettbewerbsfähigkeit zu erhalten. Eine effektive Planung von Restrukturierungen in Produktionssystemen, oft als Umzugsplanung bezeichnet, ist entscheidend, um Verzögerungen zu vermeiden und Stillstandszeiten zu minimieren. Die Identifikation von Umzugsfällen, welche den Umfang umzuziehender Ressourcen wie Maschinen, Anlagen, Ausrüstungen und Arbeitsplätzen beschreiben und die Definition zugehöriger Aufgaben sind essenzielle Schritte der Umzugsplanung. Dieser Beitrag stellt eine Vorgehensweise zur Bestimmung von Umzugsfällen mittels eines Fabrikdatenmodells vor, ergänzt durch eine Visualisierung in Virtual Reality, um Planende bei der Umzugsplanung zu unterstützen.
Control and Automation of services of the urban infrastructure offered to citizens and tourists are elementary parts of a smart city. But both rely on a stable supply of data from sensors spread across the whole city, e. g., the fill level sensors of waste bins needed for a waste management tool which we developed in a collaboration with the Regensburg city council for the on-demand collection of waste bins. Europe has a lot of historic cities like Regensburg with narrow streets and huge building walls, some made from granite and fieldstones, which often represents an insurmountable obstacle to wireless data transmission. The reduction of the road traffic volume poses an additional challenge for city planners. By means of networked planning and simulation software, the situation, state and efficiency of citywide logistic services can be monitored and optimized. In the course of such optimizations, we propose the combination of digital and logistic services. As an example, we show that monitoring state information, such as the waste bin fill levels, can be accomplished using the same vehicles and the same planning software, that is used for luggage transportation. Moreover, we describe how we adapted a solver for a variant of the TSP, namely the prize-collecting traveling salesman, to optimize the route planning dynamically.
In the realm of parallel computing, optimization plays a pivotal role in achieving efficient and scalable solutions. In this work, we present the parallelization of a hybrid genetic search for solving the Capacitated Vehicle Routing Problem with Pickup and Delivery (CVRPPD).It leverages the synergy between genetic algorithms and parallel computing to address the complex optimization problem. This hybrid algorithm combines a customized version of local search with a genetic algorithm to compute an effective solution. Our implementation makes use of the Message Passing Interface (MPI) for data distribution and parallel execution. In addition, we run multi-threaded processes on NVIDIA graphical processors using the CUDA technology, which further increases the computation speed and consequently minimizes the runtime. Parallelization also allows the best-improvement strategy to be used instead of the rst-improvement strategy while maintaining the same runtime. We store the resulting routes in a bus route database which we created as the basis of an extensive library of optimal routes for our specifc use case of optimizing bus routes in a rural area. The experimental results on real road data show that the parallel implementation of the Hybrid Genetic Search (HGS) achieves significant improvements in runtime over the sequential implementation above a certain problem size. We believe that our implementation of the parallel hybrid genetic search method can have a great in influence on optimization strategies in parallel computing and can also be applied to other subproblems of the VRP.
In this paper, we present a new approach to determine the estimated time of arrival (ETA) for bus routes using (Deep) Graph Convolutional Networks (DGCNs). In addition we use the same DGCN to detect detours within a route. In our application, a classification of routes and their underlying graph structure is performed using Graph Learning. Our model leads to a fast prediction and avoids solving the vehicle routing problem (VRP) through expensive computations. Moreover, we describe how to predict travel time for all routes using the same DGCN Model. This method makes it possible not to use a more computationally intensive approximation algorithm when determining long travel times with many intermediate stops, but to use our network for an early estimate of the quality of a route. Long travel times, in our case result from the use of a call-bus system, which must distribute many passengers among several vehicles and can take them to places without a regular stop. For a case study, the rural town of Roding in Bavaria is used. Our training data for this area results from an approximation algorithm that we implemented to optimize routes, and to generate an archive of routes of varying quality simultaneously.
