Abstract:

Virtual validation is a key pillar in developing and testing CAVs and smart transportation. However, perception sensor simulation is often treated as static. Models are validated once (if at all) and reused across projects, scenarios, and ODDs without being systematically reassessed. This results in high manual effort, limited scalability, uncertainty in development decisions, safety arguments, and virtual homologation.

This contribution presents an industrial-ready approach for continuous validation of perception sensor models. Instead of treating validation as a one-off activity, sensor model performance is continuously monitored by comparing simulation outputs against real data and predefined acceptance criteria as scenarios, environments, or model versions change. This enables early detection of validity issues and significantly reduces the risk of using simulation results beyond their validity range.

Efficient validation of simulation models is enabled by consistent use of ASAM OpenX standards, particularly the ASAM Open Simulation Interface (OSI) and OpenMATERIAL 3D. OSI provides a standardized data backbone for ground truth, sensor view, and sensor data, allowing sensor models to be reused across simulation environments without tool-specific interfaces. As a result, the validation results become tool-independent, concentrating and lowering the validation efforts to a feasible amount. Exemplary validation is shown for a lidar sensor model using real sensor data from Persival’s performance test bench.

Speaker Bio:

Philipp is the Co‑Founder and CEO of Persival GmbH, where he supports sensor manufacturers and OEMs in specifying, developing, and validating sensor models for automation, simulation, and physics‑based AI. He is an internationally recognized expert in perception-sensor simulation, serving as a founding member of the ASAM Open Simulation Interface and its Change Control Board, as well as a member of the DIN ISO 11010‑2 working group on ADAS/AD perception-sensor models. Philipp earned his PhD on metrics for specification, validation, and uncertainty prediction to ensure simulation credibility of active perception systems under the supervision of Prof. Hermann Winner in 2022. He previously contributed to major national and European R&D programs—including SET Level, VVMethods, PEGASUS, and ENABLE‑S3—where he led and advanced lidar and radar sensor‑system simulation for highly automated driving safety validation. 

Abstract: 

Modern-day Connected and Autonomous Vehicles (CAVs) represent the merger of “robotics sense-think-act in real-time” with “distributed networked” paradigms, now with the additional challenges of complexity/scale in increasingly dynamic field operations. Further, as embodiments of cyber-physical systems (CPS), the performance/value of these CAVs is derived from software-based orchestration of the underlying electromechanical hardware at component, sub-system, system, and system-of-systems levels. In this milieu, realizing long-term autonomy with CAVs within a lifecycle framework (design, analysis, refinement, prototyping, validation, and operationalization) offers unique science and technology opportunities and challenges.

The principal opportunities now lie in building upon loosely interconnected heterogeneous modular systems of systems and (re)engineering high-performance/high-confidence operational capacities in the presence of uncertainties in increasingly unstructured operational domains. New AI-enhanced paradigms are emerging to tackle the curse of dimensionality while empowering the ability to co-model, co-simulate, co-visualize, co-analyze, and co-refine intelligence algorithms (software) together with physical assets (hardware) at multiple scales through the lifecycle.

This talk will provide a high-level programmatic overview of our Clemson ConnectedAutonomy group’s efforts with Uncrewed Ground Vehicles (UGVs) capable of operating in complex and dynamic environments (on-road, off-road, and on the manufacturing shop floor). The research efforts span (i) novel data-driven learning approaches (e.g., Koopman operator theory, active learning); (ii) autonomy-oriented digital-/physical-twinning; (iii) AI-enhanced fielded autonomy; and (iv) systematic verification/validation. To support these efforts, we are also developing modular, open-architecture, open-interface physical and digital infrastructure spanning multiple scales and complexities. Vignettes from digital-/physical-twinning efforts across multi-architecture (Ackerman vs. skid-steered) and multi-scale (1/10 to full-scale) UGV platforms will be presented to showcase our efforts. URL: https://cecas.clemson.edu/armlab-cuicar/

Speaker Bio:

Prof. Venkat N. Krovi is the Michelin SmartState Chair Professor of Vehicle Automation at Clemson University – International Center for Automotive Research. His research focuses on intelligent modulation of distributed physical-power-interactions (motions/forces) between humans and autonomous systems to unlock the “power of the many." Research activities focus on the life-cycle treatment (design, modeling, analysis, control, implementation, and verification) of a new generation of ConnectedAutonomy systems for realizing Human-Autonomy synergy in emerging automotive, plant-automation, co-robotics, and defense applications. He has also taken significant leadership roles within multiple professional societies (IEEE/ASME/SAE) as well as supporting development of the 2020 US Robotics Roadmap. Further details are available from http://cecas.clemson.edu/armlab-cuicar 

Abstract: 

Winter precipitation and freeze–thaw cycles can rapidly turn road surfaces slick, creating safety-critical hazards that challenge both human drivers and automated vehicles (AVs). Yet AV driving policies and traffic operations rarely have access to reliable, network-wide, time-resolved estimates of pavement friction, especially beyond sparse point sensors and delayed maintenance reports. In this talk, we will present recent research progress toward a connected-vehicle–driven framework for friction estimation, validation, and forecasting at scale, using Pennsylvania as a statewide case study. The core idea is to leverage commercially available connected-vehicle data streams as “opportunistic probes” of roadway conditions and fuse them with complementary reference information to derive friction targets for model training and evaluation. The presentation will feature a deep learning forecasting pipeline that produces short-horizon, segment-level friction predictions to support risk-aware AV behavior (e.g., speed/headway adaptation) and proactive operations (e.g., treatment prioritization). Key takeaways include (i) how to operationalize friction inference from existing connected-vehicle telematics data, (ii) validation strategies suitable for large transportation networks, and (iii) how data-driven friction prediction can enable safer and more resilient winter AV deployment within smart transportation ecosystems.

Speaker Bio:

Dr. Xianbiao (XB) Hu is an Associate Professor of Transportation Engineering at Penn State and leads the SmartMobility Lab. Previously, he was an Assistant Professor at Missouri S&T and a founding member/director of R&D at Metropia Inc. He earned his PhD from the University of Arizona. His research includes smart mobility systems, automated vehicles, AI, and transportation electrification. He is an Associate Editor for multiple peer-reviewed journals, including IEEE T-ITS, JITS, TRR, and IJTST. He is the IEEE ITSS representative and a voting member of the IEEE Transportation Electrification Council (TEC).

Abstract:

This talk presents a practical vision for enabling autonomous parking-lot intelligence using multi-camera mapping in GPS-denied environments. It deals with problems that have been around for a long time, like unreliable localization, occlusions, low light, and the limitations of old CCTV systems. The talk introduces a scalable, vision-only framework that leverages existing camera networks to deliver real-time vehicle perception and parking management without additional hardware. The main idea of this approach is to use advanced deep learning methods, like finding objects, recognizing vehicles, and reading license plates, to keep track of the same vehicles across different camera views. A modular processing pipeline supports continuous tracking, cross-camera identity fusion, and accurate parking-slot occupancy estimation. Spatial mapping and region-aware reasoning further enhance reliability and usability. The talk also highlights real-world deployment insights, demonstrating robust performance in complex underground parking environments. Finally, it outlines a cost-effective and deployable pathway toward smarter, autonomous parking systems.

Speaker Bio:

Prof. Dhananjay Singh is a Teaching Professor at Penn State University and Director of the ReSENSE Lab, advancing intelligent solutions for smart communities. His research focuses on improving the reliability of cyber-physical systems across sectors such as healthcare, vehicles, agriculture, and IoT. With over 20 years of academic and industry experience, he has led 25+ international projects and held key leadership roles. A Senior Member of IEEE and ACM, he has delivered 100+ talks, published 200+ papers, and holds 25+ patents. He also serves in editorial and conference leadership roles and has received multiple international awards for innovation and research excellence.

Abstract:

Advances in sensing, data analysis (including AI), and connectivity have created a number of new opportunities for deploying applications and technology that can improve operations and safety. A major challenge for many of these efforts, though, is moving them from a pilot or research project stage to a larger scale implementation within a department of transportation (DOT). This presentation will discuss some recent projects in Virginia and highlight the real-world challenges that had to be addressed in moving to broader deployments. Specifically, cybersecurity, data governance policy, defining business cases and return on investment, scalability, and developing plans for ongoing operations and maintenance support will be discussed. Ways to ensure that these issues are considered early in projects in order to increase the likelihood of broader adoption will be identified.

Speaker Bio:

Mike Fontaine is the Associate Director for the Safety, Operations, and Traffic Engineering team at the Virginia Transportation Research Council (VTRC), which is the research division of the Virginia Department of Transportation (VDOT). In that role, Dr. Fontaine manages VDOT’s research program in the areas of highway safety, traffic engineering, operations, intelligent transportation systems, and connected and automated vehicles. Prior to joining VTRC, Dr. Fontaine has also worked at the Texas A&M Transportation Institute, Old Dominion University, and for the City of Charlottesville. Dr. Fontaine received his Ph.D. in Civil Engineering from the University of Virginia and is a registered professional engineer in Virginia.

Abstract:

TBA.

Speaker Bio:

Dr. Shintaro Fukushima is a Principal Researcher and engaged in research and development management in InfoTech, Toyota Motor Corporation. He is also an Associate Professor in Shiga University. He received his B.S. and M.S., and Ph.D. degree in information science and technology (machine learning and data science) from The University of Tokyo. Currently, he leads research and development in InfoTech, in particular, artificial intelligence (AI), data science, and the associated data engineering. His work includes AD (advanced driving) / ADAS (advanced driver-assistance systems), intelligent cockpit with generative AI and agentic AI, privacy-preserving machine learning methods such as federated learning, intelligent transportation systems, and data analysis for electric vehicles, factories, power-train, and material science. He has served as a committee member of the AI Quality Management Study Committee in The National Institute of Advanced Industrial Science and Technology since 2019. He also

served as an editor of the Editorial Board of Bulletin in The Japan Society for Industrial and Applied Mathematics from 2021 to 2024.

Abstract:

The use of Generative AI (GenAI) is seemingly having positive effects on further automating software engineering processes even beyond code and test generation. Typical activities for software engineers are substantially being redefined, and expertise on system architecture, orchestration, and large-scale refactoring are gradually becoming routine daily activities in the context of agentic software engineering. In this talk, the speaker will share selected highlights and critical reflections on the benefits and challenges of agentic engineering from ongoing research activities in the area of automotive software engineering.

Speaker Bio:

Dr. Christian Berger is Full Professor at the Department of Computer Science and Engineering at the University of Gothenburg, Sweden, and received his Ph.D. degree from RWTH Aachen University, Germany, in 2010. He coordinated the research project for the self-driving vehicle "Caroline," which participated in the 2007 DARPA Urban Challenge Final—the world's first urban robot race. He also co-led the Chalmers Truck Team during the 2016 Grand Cooperative Driving Challenge (GCDC) and is the leading software architect at Chalmers Revere, the laboratory for automotive-related research. His research expertise is on architecting cloud-enabled cyber-physical systems and event identification in multi-modal, large-scale, time-series datasets to support the engineering of safe and trustworthy AI-enabled systems in the era of agentic engineering.