Marc Haubenstock

Code the Future of Navigation: Students with ML (Machine Learning) and LLMs (Large Language Models) approaches in the fast lane! 🚗 Creativity, teamwork, and vision were the driving forces behind our hackathon with the Karlsruher Institut für Technologie (KIT). Twenty-five talented students gathered at our location in Karlsruhe to revolutionize navigation. Our "Code the Future of Navigation" hackathon set the stage for creative exchange between research and industry, with three exciting focus areas: ✅ Personalized route planning: Adaptive navigation tailored to individual preferences and contextual factors. ✅ Driver behavior analysis: Creating dynamic driver profiles based on behavioral data. ✅ Routine & pattern recognition: Predicting driving behavior to enhance navigation efficiency. Different teams – diverse innovative approaches: 💡 Predictive route modeling – Using ML methods like random forests to predict routing decisions based on historical data. 💡 Utility-based decision modeling – Combining ML with utility theory to analyze and weigh key factors influencing route choices. 💡 Carpooling & personalized travel patterns – Identifying and leveraging individual mobility patterns to optimize shared rides and route planning. 💡 Weather-aware route optimization – Integrating weather data into routing algorithms, including meteorological data processing and optimization strategy adjustments. 💡 Personalized stop recommendations with LLMs – Leveraging LLMs to generate customized stop recommendations based on driver profiles and contextual factors. Inspiring talks from experts like Jens Nahm, Harry Baur, Peter Dung, Franziska Pöpperl, Alexandra Wins, Daniel Saad, Johannes Nonnenmacher, and Jochen Mehlhorn fueled the spirit of innovation. Special thanks to the KIT and Fabian Rybinski for supporting the collaboration between academia and industry.   To support these innovations, we provided high-quality synthetic datasets simulating real-world driving behavior, generated using data from our business stream “One Customer”, LLMs, utility theory, and probabilistic methods. The results were impressive! They will be integrated into ongoing projects and presented to our Mercedes-Benz AG Business Partners. Our hackathon was a huge success, marking a significant step towards a connected and intelligent future of navigation. The collaboration between students and our experts showcased the immense potential of merging research and industry. 🤝 #techinnovators #machinelearning 

170

Our researcher Wilma Schmidt is currently attending the 𝗘𝗦𝗪𝗖 𝟮𝟬𝟮𝟱 Conference in Portorož, Slovenia, to present our latest work in the In-Use Track. The paper, titled "LLM-Supported Mapping Generation for Semantic Manufacturing Treasure Hunting," explores how large language models (LLMs) can support the creation and maintenance of knowledge graphs in complex manufacturing environments. 📊 In companies like Bosch, simple inventory questions can turn into major data challenges due to distributed systems and heterogeneous sources. Knowledge graphs offer a promising solution, but generating and maintaining them requires significant manual effort. Our approach uses LLMs to semi-automatically generate YARRRML mappings and applies ontology reduction methods to overcome token limitations when dealing with extensive manufacturing vocabularies. This work highlights how combining LLMs and KGs can enable more scalable and efficient semantic integration of structured data in industrial contexts. This paper is the result of a great collaboration between Bosch Research — Evgeny Kharlamov, Irlan G., and Wilma Schmidt — Electrified Motion with Dr. Lena Wagner and Tobias Huschle from Robert Bosch GmbH in Bühl, Germany, and Adrian Paschke from Freie Universität Berlin and the Data Analytics Center (DANA) at Fraunhofer Institute FOKUS. Congrats to the whole team on your impactful work! 👏 #BCAI #BoschResearch #KnowledgeGraphs #LLM #SemanticWeb

66

2 Kommentare

🎓𝗔𝘀 𝗮 RWTH Aachen University 𝗮𝗹𝘂𝗺𝗻𝘂𝘀, 𝗜 𝗮𝗺 𝗽𝗿𝗼𝘂𝗱 𝘁𝗼 𝗯𝗲 𝗳𝗲𝗮𝘁𝘂𝗿𝗲𝗱 𝗼𝗻 𝘁𝗵𝗲 𝘄𝗲𝗯𝘀𝗶𝘁𝗲 𝗳𝗼𝗿 𝗺𝘆 𝗱𝗲𝗴𝗿𝗲𝗲 𝗽𝗿𝗼𝗴𝗿𝗮𝗺, 𝗖𝗼𝗺𝗽𝘂𝘁𝗮𝘁𝗶𝗼𝗻𝗮𝗹 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 𝗦𝗰𝗶𝗲𝗻𝗰𝗲! If you don't know what to study (or your kid is unsure), I can highly recommend such interdisciplinary programs. The combination of mechanical systems, numerical mathematics, and high-performance computing is a strong enabler for all kinds of jobs in data science, simulation, AI, software development, etc., etc.! Back when I was still a student, I was a member of the audit committee for CES, where we already started discussing how to make the program easier to understand and more attractive to young people. After all, it's a lot and can be off-putting for many students after graduating from high school. Thus, I am super excited to see that the team around Manuel Torrilhon, Marek Behr, and Uwe Naumann put so much effort into this new web presence of the degree program. Thank you, I really like it! And great to see Maike Scholtes from my cohort back then represented as well :) In my current role as a research scientist working on autonomous driving at FZI Research Center for Information Technology, I benefit from that education a lot. After all, vehicles are mechanical systems, AI is basically math, and both training and inference require high-performance compute solutions. If I could go back in time, I would study CES at RWTH again! You can find my whole little quote on the Computational Engineering Science website: https://lnkd.in/eye7_RMx #education #degree #bsc #msc #university #school #alumni #ces #student

88

2 Kommentare

🚗💡 Successful Completion of the Urban Digital Twins Project from the Institut für Stadtbauwesen und Stadtverkehr der RWTH Aachen University, Design Computation - RWTH Aachen University  and Geodätisches Institut und Lehrstuhl für Bauinformatik & Geoinformationssysteme - RWTH Aachen with BMW Group!   Our joint project exploring the potential of Urban Digital Twins (UDTs) has been successfully completed! The aim was to develop innovative use cases that demonstrate how urban data can be utilized more efficiently – particularly by integrating vehicle data as mobile sensors.   ☝️ Why digital twins? UDTs empower cities to enhance decision-making in areas like traffic management and urban development. They can improve operational efficiency, reduce costs, and contribute to better environmental and sustainability outcomes. 🔍 Project Overview: – A comprehensive analysis of around 80 cities worldwide was conducted to identify best practices and potential use cases. – Interviews with representatives from selected cities helped refine ideas and evaluate possible scenarios. – Co-creative workshops, including on-site workshop in Leipzig, enabled discussions and advancement of concrete pilot projects.   🚀 Results: The project highlights the immense potential of UDTs, particularly through the intelligent use of vehicle data. This technology has the capacity to deliver significant contributions to the urban development of the future. Many thanks to everyone involved in this success – especially Jochen Reuß, Dr. Sascha von Behren, Felix Przioda, Uwe Döring und Gregor Schley of the BMW Group, Mathias Boedecker, Mirko Mühlpfort, Sebastian Graetz, Timon Dischmann and Diana Zschörnig from the City of Leipzig, as well as our interview partners Günter Koren, Reinhard Odrei, Jürg Lüthy, François Ruettimann, Eric Jeansson, German Castignani, Roland van der Heijden and Terence Tan.   #DigitalTwins #Innovation #SmartCities #BMW #UrbanDevelopment #RWTH

73

3 Kommentare

My research on using AI to monitor glacier retreat was recently featured by FOCUS 🧊🛰️ The article presents how deep learning methods can be applied to radar satellite imagery to automatically detect changes in marine-terminating glaciers 🌍, contributing to more efficient large-scale glacier monitoring through “green AI” 🤖💚 Thanks to the Pattern Recognition Lab and the M3OCCA international doctorate program for their continuous support 🙌 Read the full article here: 🔗 https://lnkd.in/dPzgpmaV #GreenAI #AI4Earth #AI4EO

133

5 Kommentare

#TEC25 Sylvia Schwaag Serger, co-author of the Heitor Report and Deputy Chair of Austria‘s Research Council FORWIT – Rat für Forschung, Wissenschaft, Innovation und Technologieentwicklung (apart from her main job(s)) at #TechnologyTalksAustria: „Technology is driving geopolitics.“ „Our notion of competitiveness should really be expanded beyond a purely economic concept.“ 💯 Bundesministerium für Innovation, Mobilität und Infrastruktur

126

7 Kommentare