Ralf Herrtwich

As we, at ZF Passive Safety Systems together with Dr. Michael Eisenbarth, gear up for the 34. Deutscher Montage­kongress - Let's dive deep on industrial AI and its impact on productivity. The recent study by the IW Medien GmbH (Institut der deutschen Wirtschaft Köln Medien GmbH) predicts an impressive €330 billion additional value through generative AI, provided that at least half of our companies embrace these transformative technologies. 🤝 Let’s Connect! If you’re attending the congress (link to register in the comments) and you should - Feel free to reach out here on LinkedIn or find me during the sessions. Let’s exchange ideas, collaborate, and discuss the future of assembly processes and industrial AI. 🔍 Key Questions We’ll Explore at the Congress: How can we profitably integrate machine learning and deep learning methods into assembly processes? What strategies can optimize existing assembly resources using AI? Can generative AI truly save an average of 100 hours per employee per year in assembly, as the above mentioned study suggests? Is AI the solution to our skilled workforce shortage? And how do we ensure a seamless transition for our assembly teams? #AI #Productivity #AssemblyOptimization #CongressNetworking #ZFambassador ZF Group #industrialAI

21

1 Kommentar

We also discuss this briefly in our upcoming publication: coming soon, Gallun, S., and Loftus-Otway, #Legal Research Digest 91, prepared under NCHRP 26-03. Multistate Coordination and Harmonization For Automated Vehicle Legislation.  Submitted August 2023, TRB, Washington DC.(legal research digest that identifies the legal issues and barriers that exist for the national harmonization of state AV laws.) https://lnkd.in/g3eV5U_5 #automatedvehicles hopefully out by July

2

Very interesting as we discuss in our Group the same what Happens when AI has Input data that is Beyond training data. Is there a chance to close the gap in a safe way? What is safety the moment you detect the gap? And last- can a model or procedure of AI be better than the people who implemented the training algorithm?

4

Really interesting article and how race cars can be used to further the development of autonomous vehicles!

13

BMW's global vehicle sales for the first six months of 2024 have been published by #BMWBlog. As the analysis on the left of the picture shows, by using adoption of the curved display for each model type as a proxy for #driver #monitoring #system fitment, we can approximate #DMS units at BMW Group. For the 2025 model year, only the 8-Series, #X4 and #Z4 are judged to not feature the curved display, with the M5 having been withdrawn. The exact start of production date for each model upgraded to the curved display for the 2025 model year is unknown, but BMW looks to be on course to hit 500,000 vehicles a quarter with DMS during this calendar year, and may reach that milestone as early as the July-September (current) quarter. We will know more next week when Seeing Machines publishes numbers for quarterly cars on road for the three months covering April-June. Smart Eye releases its financial results the week after, but it has never published quarterly cars on road data. BMW is judged to have dropped Smart Eye as its DMS supplier and moved to Seeing Machines DMS for the Gen2 product. BMW has standardized on DMS integration within the instrument cluster, and is assumed to be using a #cockpit #domain #controller running Operating System 8/8.5/9 on the third-generation Qualcomm cockpit #applicationsprocessor. Change is coming in 2025, with the introduction of the BMW #NeueKlasse platform which will run Operating System 10 in the cockpit. The Neue Klasse will feature BMW #PanoramicVision, and a 3D head-up display #HUD. The tier-1 is assumed to be Valeo, with Seeing Machines set to continue as the DMS provider going into the third-generation product. The May 2024 edition of Semicast's market report "ADAS & Automotive Vision-based Driver Monitoring Systems" is available for purchase. Significant segmentation of the DMS/OMS market is provided. This is summarized as follows: - by movement detection into "Owl-only" and "Owl & Lizard" detection - by feature set - by processor type - by CMOS image sensor location - by CMOS image sensor resolution - by driver and occupant monitoring - by front-seat and whole-cabin monitoring Message me for a copy of the sales brochure and sample pages for the study. My inbox is open. BMW Blog link: https://lnkd.in/g2uNMJd3

39

Our advisor and professor for automotive technology Markus Lienkamp has spoken about the current state of technology and deployments in autonomous driving with automotiveIT. Some of the key insights he shared: ✅ There is a multitude of sensing technology. Its going to be essential to have the right mix between redundancy, cost, and performance. Focusing only on a single technology is most likely not going to be succesful. ✅ The dependence on high-definition maps remains a major bottleneck and technological challenge. Mapless approaches are a promising alternative and should be pushed forward. ✅ Connectivity via V2X will not solve the safety problems in mixed-traffic scenarios. All autonomous vehicles are going to need the required perception and sensor-fusion technology to operate reliably and safely without them. Read more in the full article: https://lnkd.in/dfJ7Y3Zm (german only) #autonomousvehicles #autonomoustrucking

16

4 Kommentare

SW Issues at VW Summary Volkswagen Group's software division, Cariad, is facing significant delays, pushing the launch of the Scalable Systems Platform (SSP) from 2024 to an uncertain date after 2026. Audi, initially set to launch an SSP-based model in 2026, now expects delays until 2027 or 2028 and has considered purchasing a platform from a Chinese automaker. The Volkswagen brand is also affected, with next-generation models like the electric Golf and ID.4 pushed back to 2029. Delays have also impacted the launch of the large electric Volkswagen SUV T-Sport, now moved from 2028 to 2031. (Link to article: https://lnkd.in/drju5Cd8) My take: 1. Creating a SW platform for multiple models always results in a overly bloated "superset" of a platform that ends up being a beast of its own. This is a very dangerous direction especially when the effects of it will be felt across multiple OEM brands, countries and years to come. 2. The only way these kind of initiatives end up working is to start simple and small. The platform has to become the thinnest minimum viable piece of SW that is really robust and smart. And building from there. 3. Moving all the SW expertise outside of the OEM is also a bad move. SW abstraction layers, no matter how clean they can be split on a diagram, will have a lot of dependencies to each other. And the only way to figure all that out, especially when it is the first iteration is to keep the right engineers together and focused on a common goal. Starting a new SW company does not necessarily make that easier.

61

6 Kommentare

Understanding objective and subjective reasons for the preferred usage of a private car is key for developing demand-oriented (and thus succesful) multimodal mobility alternatives. This is why we work together with academia in order to find out what drives mobility behaviour and which needs need to be adressed in order to develop the right mobility solutions that improve the mobility experience of car drivers and multi-mobilists.

15

🚀Breaking News! 🚗 "Revolutionizing Traffic Flow: The latest research paper, 'Research on Dual-Clutch Intelligent Vehicle Infrastructure Cooperative Control Based on System Delay Prediction of Two-Lane Highway On-Ramp Merging Area,' tackles a common bottleneck in modern highways. 🔍 This paper present a groundbreaking dual-clutch cooperative planning model, inspired by dual-clutch transmission, specifically designed for two-lane highway scenarios. 🚧 This model, coupled with a cutting-edge system delay prediction model utilizing adaptive Kalman filter, elitist genetic algorithm (inspired by imitation learning), and RBF neural network, ensures seamless merging control even under communication delays. 🔄 Simulation results reveal DPDM's (Delay Predicted Dual-Clutch Merging Control Model) exceptional potential in boosting safety, expediting merging, optimizing flow speed, and reducing fuel consumption. 🌿 A must-read for researchers, engineers, and policymakers in the field of connected & automated vehicles! 📚 Click here to check the paper: https://lnkd.in/g_2YTmQf   #DualClutchControl #CAVs #IntelligentInfrastructure #TrafficFlowOptimization #FuelEfficiency"

3

Thought of the day: "United we stand divided we fall" People working together are more likely to be successful (effective) than those who work individually. Same is true for ADAS systems like AEBS, LDW & DMS. If these systems work together, they can handle many more scenarios and can be used to improve the overall performance & effective ness of the system. For example, driver distraction cases like driver talking on phone or driver looking sideways or towards the infotainment screen even though dangerous may not always be disastrous (like on empty stretch of road with a controlled within lane driving) , the situation will only escalate if obstacle is detected on collision path or lane deviation is detected Under such circumstances the distraction threshold and duration both can be slightly relaxed (to avoid driver irritation) if no danger is perceived for next 3 seconds by other systems like #FCWS / AEBS and #LDW. Unfortunately, #DMS as a standalone system without the knowledge of on road conditions will result into unnecessary irritation to driver though distraction warnings as soon as his eyes are off the road Similarly, if the driver is drowsy and even though #DMS sounds an alarm, it will take some time for the driver to wake up and get into his full senses, unfortunately this delay could be good enough to cause an accident, this situation could be easily handled through #AEBS by applying brakes or active #DMS which cuts off accelerator pedal This is the main reason why at Starkenn Technologies we have developed a 360 Degrees safety system, where all the systems work together to reduce accidents by alerting driver and also taking proactive actions like accelerator cut off / braking and at the same time avoid unnecessary alarms which might irritate driver (which may eventually result into switching off the safety system(s) by driver)

17

What’s next for software-defined vehicles? What will modern architectures - often referred to as operating systems - look like? And how is KPIT positioning itself in Europe - and globally - to remain a key partner in this area for car companies? Together with KPIT’s co-founder and CEO, Kishor Patil, I talked to Automobil-Elektronik (all-electronics.de) about our strategy, our growth ambitions and how we help mobility companies develop architectures for software-defined vehicles faster and with greater flexibility. When I joined KPIT last year, I did so because I was impressed by the company’s ability to think ahead of the industry curve. That’s only possible when you are 100 percent industry-focused as we are. KPIT is not only providing specialized automotive software solutions. As a partner, our aim is to enable the business for our clients - by bringing the architectures of future car generations to life, making them secure, and putting technology at the heart of our clients’ business.  If you want to learn more about our approach to software-defined vehicles and why Europe continues to be a growth market for us, read the full interview. German: https://lnkd.in/d45qWDrX  English: https://lnkd.in/dhZTDtSx #sdv #softwaredefinedvehicle #readyforgrowth

222

4 Kommentare

"The concept of Safety of the Intended Functionality (SOTIF) is crucial in the realm of environmental perception, particularly in autonomous driving systems. The genesis of SOTIF can be traced back to a significant accident in 2016 involving a Tesla Model S on a U.S. highway. The vehicle, traveling at 119 km/h, collided with a trailer of a crossing truck. Tesla reported that the radar sensor mistook the trailer for a bridge (which it considered underivable), and the camera failed to detect the white trailer against a bright sky backdrop. Consequently, the vehicle did not stop, resulting in a fatal crash." Dominik Seidel. Read more about SOTIF in context of Environmental Perception in the upcoming Revision 1.2. Here is another example of a Tesla Crash, similar to the scenario, mentioned above: https://lnkd.in/ecBeBKC6

11

1 Kommentar

If you are working in ADAS developments, don't miss this opportunity. Get detailed insights into methodologies for capturing and analyzing data from various ADAS sensors and ECUs. #ADAS #Webinar #Vector #DataLogging #DataAcquisition #AutomotiveTech #VehicleSafety #AdvancedDriving #SmartMobility #AutomotiveEngineering

14

Advanced Driver Assistance Systems Features: ADAS Explained 👇 https://lnkd.in/eGW3Y9um

2

Posting this for my non-automotive family, friends, and contacts, even though these levels were announced 10 years ago (published by SAE International as J3016) . Courtesy: Swiss Transit Lab (Original post was in German) https://lnkd.in/gp-Ggmxe . "🚗 Automation Level – The 5 Steps to a Self-Driving Vehicle 🚘 Level 0 – No automation: The driver has complete control over all aspects of driving. 👣 Level 1 – Feet off: Semi-automated systems such as cruise control or lane departure warning support the driver in individual tasks, but the driver remains active at the wheel and must be able to intervene at any time. 🙌 Level 2 – Hands off: The vehicle can control multiple driving functions at the same time (e.g., accelerating, braking, and steering), but the driver still needs to monitor the environment and be ready to intervene. 👀 Level 3 – Eyes off: Here, the vehicle takes complete control of the driving under certain conditions, but the driver must be able to intervene if the system requests it. 🤯 Level 4 – Brain gone: The vehicle can drive fully automatically in certain environments without the intervention of an operator. The "driver" becomes a passenger. 🤖 Level 5 – Full Automation: This is the ultimate goal! The vehicle can drive autonomously at any time and in all conditions. It masters all tasks independently, but has neither a steering wheel nor pedals, depending on the case. "

3

1 Kommentar

Autonomy beyond ADAS Series 🚗 Let's do a deep dive into technical intricacies & a brief comparison of ADAS & Autonomy beyond ADAS (going from AV1.0 to AV2.0). You'll see how advancements in Foundational AI Models are overcoming limitations of prior approaches and granting more freedom to achieve Autonomy. It is very exciting to see how AI is shaping the next automotive revolution. At Minus Zero, we are building Foundational AI Models for Driving. Stay tuned for more updates!🚀 #minuszero #ai #technology #adas #autonomousvehicle #india #innovation

107

1 Kommentar