leveLXData by fka

🚗 New Paper: Enhancing Autonomous Vehicle Testing with Better Scenario Descriptions Have you worked on the description of scenarios, hoping for easier and more efficient ways? Nicola Kolb, Florian Huber and Alexander Pretschner from Technische Universität München published a new paper, "Automatically Improving Scenario Descriptions Derived From Recorded Traffic", presenting a novel approach to improve scenario descriptions—a crucial step in testing autonomous vehicles for safety and performance. As scenario-based testing evolves, innovations like this brings us closer to safer and more robust self-driving systems. Key Contributions of the Study: ◾ Uses real-world traffic data to automatically generate and refine scenario descriptions ◾ Measures scenario description quality using the Fréchet distance, improving accuracy ◾ Optimizes inadequate descriptions using a two-step process, adjusting maneuvers and semantics ◾ Applied to 843 intersection scenarios from the #inD dataset, the reconstruction quality is doubled Why This Matters: Accurate #scenario descriptions are fundamental in autonomous vehicle testing, helping simulate real-world behavior more effectively. This research automates the improvement of these descriptions, saving time and enhancing the reliability of #AV #safety testing. The Role of #inD in Advancing AV Testing: By leveraging high-resolution traffic recordings, the research was able to validate and enhance the quality of automatically generated scenario descriptions, making them far more realistic for autonomous vehicle testing. Congratulations to the authors for their contribution! You can find links to the paper and the #inD dataset in the comments below. #Scenarios #ScenarioDescription #inD #leveLXData #Testing

🌨 There is no such thing as bad weather! (for data collection) To get a full perspective on #naturalistic driving it's crucial to take different environment conditions into account. This is why our #leveLXData team is constantly advancing our computer vision and data processing pipeline to cover more and more adverse weather conditions. Fog, snow, rain,... get our high quality data to check and compare the impact of reduced visibility on #driving #behavior. Accounting for these differences is crucial for developing safe and reliable #ADAS and #AutomatedDriving systems. Curious to learn more about the details in our data? Reach out to Thomas Keutgens and Christoph Klas! #ScenarioData #AutonomousDriving #TrafficSafety #FogConditions #AdverseWeather #OpenSCENARIO #OpenDRIVE

Proud to Support the STADT:up Project! We're excited to see our data being used by Johannes Lindner, PhD student at the Technische Universität München, TUM - Lehrstuhl für Verkehrstechnik, as part of the EU and BMWK-funded project STADT:up. Johannes focuses on #modeling pedestrians and cyclists (VRUs) for microscopic traffic simulations and synthetic scenario creation for software-in-the-loop testing. By using our high-precision real-world trajectory data of vehicles and VRUs, leveLXData provides a ready-to-use foundation from complex urban locations with a high amount of interaction between vehicles and VRUs that can be used to make simulation more realistic. 🔍 About STADT:up: This publicly funded project aims to develop continuous automated driving and sustainable, intermodal mobility concepts for cities. This requires among others the reliable detection and modelling of all traffic participants in the urban domain. Base on 5 subprojects the overall goal of the project is the creation of a functional automated driving prototyp working in challeging inner-city traffic scenarios. We’re proud that our data contributes to automated urban transport and a looking forward to the results of the project! Want to learn more? Check out the project here: https://lnkd.in/e2T8scQi #leveLXData #TrafficData #TrajecotryData #STADTup #AutomatedDriving #PedestrianModelling #CyclistModelling

🚗🛵 #Heterogeneous Highway Trajectory Data from #Asia Studying heterogeneous traffic and developing and testing ADAS and AD functions tailored for heterogeneous traffic is becoming increasingly important. For this reason, we have recorded motorway data in heavily populated Hong Kong. What is special about their motorway traffic is the significantly higher proportion of #motorcycles compared to U.S. and Europe highways. The data captures #traffic #behavior and #scenarios in one of the world’s most densely populated areas:   ◾ Collected over 14 hours at 5 locations   ◾ Highways with entries and exits and high number of interaction ◾ Almost 100.000 road users tracked, including a significant number of motorcycles, enabling insightful comparisons to Europe or the U.S. This dataset provides unique advantages for researchers and developers working on ADAS/AD systems tailored for busy, diverse traffic environments. Whether you're exploring driving patterns in Asia or want to adapt your driving behavior to this type of traffic, this is the data you need!  Contact Christoph Klas & Thomas Keutgens for more details.  #leveLXData #ScenarioTesting #DriverModelling #ADAS #AutomatedDriving #Mobility #TrafficData #Validation

⛽ New Paper: How Eco Driving Can Benefit from Real-World Drone-based Trajectory Data We’re excited to share an insightful new study, "Incorporating driving behavior into vehicle fuel consumption prediction: methodology development and testing" by Huthaifa I. Ashqar, Ph.D. from Arab American University and Columbia University together with Mahmoud Obaid, Ahmed Jaber, Ph.D., Rashed Isam Ashqar, Ph.D., Nour Khanfar, Mohammed Elhenawy. The study leverages our #inD dataset to derive a methodolgy to describe the relationship between driving styles and fuel efficiency. ◾ Research Problem: Fuel consumption is deeply influenced by driving behavior, but accurately modeling this relationship in dynamic, mixed-traffic environments has remained a challenge. This study aimed to develop a comprehensive framework that incorporates driving patterns into fuel prediction models for better ecological and economic driving solutions. ◾ Approach: The study used the #inD dataset, which includes trajectories of over 13,000 road users in urban German settings. The researchers combined different machine learning models such as linear regression and Random Forest algorithms for driving style classification and a microscopic fuel consumption model. ◾ Key Findings: Aggressive and conservative driving lead to higher fuel consumption compared to normal driving styles. The Random Forest model achieved a remarkable predictive accuracy (R² = 0.956), highlighting the effectiveness of machine learning for such analyses. Insights from the study can inform Advanced Driver Assistance Systems (ADAS), enabling real-time fuel-efficient driving strategies. Congratulations to the authors for their contribution to making vehicles more sustainable! You can find links to the paper and the #inD dataset in the comments below. #FuelEfficiency #DrivingBehavior #MachineLearning #inD #ADAS #TrajectoryData #UrbanDriving

🍀 Interchange Merging - How would an automated system perform? In our #LXD database, we can filter trajectories for specific #scenario types. For example, to analyze human #behavior at cloverleaf interchanges and formulate requirements for an #automated #driving #system to handle such typical yet complex interactions. In our analyses of these scenarios, we found, that a typical lane change parameterization of an automated system will not be suitable for this case. The #interaction behavior needs to be adapted with regard to accepted headway times and distances and it needs to account for cooperation. To systematically examine big amount of such scenarios, we export and provide the data into individual scenario snippets based on #ASAM #OpenSCENARIO and #OpenDRIVE. With scenarios like this, our data can help developers create smarter, safer systems that understand and predict complex traffic interactions. Want to dive deeper? Reach out to explore our public and commercial datasets! 🚦 #leveLXData #TrafficData #TrajectoryData #ADAS #AutonomousDriving #ScenarioBasedTesting #ScenarioData 

📝🚗 Paper Highlight: A unified probabilistic approach to traffic conflict detection! So great to see how our #highD dataset continuously helps researchers for their innovative studies! In the recently published paper "A Unified Probabilistic Approach to Traffic Conflict Detection", authors Yiru(艺茹) Jiao(焦), Simeon Calvert, Sander van Cranenburgh, and Hans Van Lint from ADaS Lab and Delft University of Technology presented an innovative framework for traffic conflict detection. Challenge: Detecting traffic conflicts consistently across diverse road environments and scenarios is essential to ensure safety in complex traffic systems, especially with autonomous vehicles. Existing methods often struggle with adaptability and inconsistencies. Solution: The team developed a unified probabilistic framework that models traffic conflicts as extreme events in road interactions. By using statistical learning tasks, their approach evaluates collision risks across varying contexts while maintaining consistency and explainability. The approach is conflict-label-free and interlinked, with the output of one being the input of the next task, thus can be further embedded into "end-to-end" autonomous driving. For analysing their concept they used our #highD dataset, collected on German highways, which provided the researchers with high-precision #naturalistic #traffic #data. Its ability to capture detailed vehicle interactions and lane changes enabled robust model training and testing under real-world conditions. Congratulations to the authors for their contribution to traffic safety research! You can find links to the paper, their code and the #highD dataset in the comments below. #leveLXData #highD #TrafficSafety #Innovation #AutonomousVehicles #Research #ConflicDetection #EdgeCaseDetection

🎄 Christmas Party with the leveLXData #Team! The fka GmbH and Institut für Kraftfahrzeuge (ika) - RWTH Aachen University Christmas Party was the perfect opportunity to celebrate the amazing leveLXData team and everything we’ve accomplished together this year! A huge thank you to everyone for your incredible hard work, dedication, and creativity in pushing the boundaries of what we can do with high-quality naturalistic traffic and scenario data! #leveLXData #ChristmasParty #Teamwork #Innovation #TrafficData #ADAS #AutonomousDriving #DroneData #ScenarioData

🎄 2024 has been quite busy for #leveLXData's team   We have added lots of new data to our database, worked in really exciting projects with new and old partners, implemented nerdy stuff to get more info out of our data, improved our pipeline efficiency and had countless productive talks.   For 2025, we have numerous exciting ideas and innovative features on the way! Stay tuned 🚀 Thank you to all of #LXD’s friends, partners, and customers. Wishing you a relaxing, joyful Christmas break and a great start to 2025!

🔍 LXD @ ASAM International Conference 2024 This year’s ASAM International Conference was a great success! A big thank you to Marius Dupuis and the entire ASAM e.V. - Association for Standardization of Automation and Measuring Systems team for organizing such a fantastic event and bringing together all these renowned expert. It's been a privilege to be part of the discussions shaping the future of mobility. We were happy to give some insights about our innovations in our poster, "From Real-World Data to Requirements: Scenario-based Specification for ADAS". It highlighted how naturalistic traffic data can make specifying ADAS more efficient and ensure safer, smarter systems. Our Key Takeaways from the conference: - Large Language Models for Scenario Creation: Tools like generative AI are paving the way for creating realistic, complex scenarios for validation (e.g., by leveraging #OpenSCENARIO and #OpenDRIVE formats). - Tille Karoline Rupp (Porsche Engineering), Zuqiu Mao (51Sim), Yves Peirsman (Deontic) - Importance of Real-World Data: Scenario extraction from real-world data remains critical for validating autonomous systems. Thorsten Püschl (dSPACE), Thomas Unger (Verkehrsunfallforschung an der TU Dresden GmbH), Jens R. Ziehn (Fraunhofer IOSB) - Advancements in OpenODD and OpenLABEL: New standards are driving improvements in defining and tagging operational design domains (ODD) and enhancing scenario-based validation pipelines. - Jens R. Ziehn (Fraunhofer IOSB), Heiko Scharke & Tahir Eren Mungan (AVL) We took many inspiring ideas with us and can't wait to start implementing them into our solutions. We are convinced that collaboration and the usage of universal standards are the best way to promote the development of safer automated driving systems. 🚀We're excited to see what upcoming projects, technologies and standards will set the tone for the next conference in 2 years! #AIC2024 #scenarios #automateddriving #ADAS #smartmobility #driverbehavior #TrafficData #ScenarioData