The third and penultimate part of the Automotive Tech.AD Diary 2017 comes with two presentations from car manufacturers. General Motors and Audi are a solid part of the Automotive Tech.AD community and once again showcased their progress on the road to autonomous driving during the two conference days.
GM: Sensing and perception technologies for automated driving
Dr. Wende Zhang from General Motors gave insights into the company’s work from conception to production of sensing and perception technologies. He names four pillars that build the motivation to develop the technology:
Yesterday, today, tomorrow
Together with Carnegie Mellon, GM already involved in the development in the 1990s and was also participating in the two DARPA Challenges. Back then no vehicle was able to reach the goal in the course of the first DARPA Challenge. But in 2010 the company was part of VisLab, an autonomous journey from Italy to China. In the same year they presented the urban autonomous concept car EN-V at the Shanghai Expo.
Today GM is among the leaders of the development, together with BMW, Audi, Daimler, closely followed by Ford, Volvo and Toyota. More close competitors at this stage are Honda, JLR, Hyundai (KIA), Nissan, Telsa, Volkswagen and Fiat Chrysler. Mazda, Renault, PSA and Mitsubishi are considered challengers. In the recent years the willingness to cooperate significantly rose – GM for example closed new partnerships with Telogis, Fline, Lyr, Cruise and Side-Car to forward the development of autonomous driving technologies. Dr. Zhang stated that GM is preparing itself for the rise of highly automated on-demand vehicles and, of course, fully autonomous driving.
From sensors to the drive
Automated driving is not possible without sensor technology. Moving objects like cars, cyclists and pedestrians have to be recognized as well as curbs and road markings in order to maneuver the vehicle automatically. Whilst doing so, the driver needs to be observed to assess his current awareness. The data collected by cameras and LiDAR systems is summarized to adjust GPS data. This enables the car to stay in the center of the lane.
Dr. Zhang names the cascade as one method to recognize road markings takes account of the regional characteristics and helps to reduce the false alarms. Road boundaries on the motorway are analyzed using the Structured Hogh Votin method developed by Canegie Mellon University. Together the motion and roadside analysis fusion provides the data to recognize pedestrians. In contrast to the traditional approach the motion analysis of the GM system realizes if the road is free or not. This is the result of image processing combined with modeling and respective algorithms. The driver’s monitoring scheme also follows a cascade. The face is captured automatically. If the system loses the driver’s eyes, the tracker is re-initiated. Head movement patterns are memorized to better evaluate and follow the next actions of the driver.
In his outlook for the future Dr. Zhang again pointed out the importance of cooperations and hard- and software improvements. Sensors still need to improve in order to reduce the costs, facilitate installations and avoid malfunctions. However he says: “Autonomous driving is not a mission-impossible”.
Audi keynote on automated parking in parking garages
One of the highlights of the second conference day was focused on automated parking – Florian Schuller, Leader of the project Piloted Parking by Audi, outlined the impact of automated parking in parking garages on functional architectures & business implications.
Easing the drivers’ pain
For many drivers parking is a quite annoying affair. The parking automation approach could be their remedy, at least regarding parking in car parks. So far people got to know optical or steering assistance during the parking process enabled by the convenient ultrasound sensor. Meanwhile sensor technology has been refined with camera systems and data fusion procedures. Mr. Schuller pointed out, that it is possible to fully automate the parking process in car parks in the future. But for this you not only need high-performance sensors, but the appropriate infrastructure and a legal framework at the national and international level.
Ecosystem: Autonomous parking
If car manufacturers and customers jumped on the autonomous parking train, the next task would be the application of a parking garage management system which allows drivers to pay from the vehicle. Audi for example is running a cloud service providing the customer with an app. But that is only one approach – with parking automation companies can access new business fields on an existing market. This also applies for the companies operating a car park. The possibility to park autonomously makes the parks more attractive for car drivers. With automated parking services the operators can handle capacity more efficiently and prevent traffic congestions.
Infrastructure vs. onboard systems
There are a lot of possible application fields for autonomous parking but it is a long-term investment. According to Mr. Schuller the concepts lies in an area of conflict between vehicle intelligence and infrastructure development. We will still have a coexistence of manual and automated vehicles for a long time, so it will be vital to have intelligent system within vehicles and infrastructure that are able to distinguish between them.
Analogous to aviation Audi is thinking of 3 main strategies to manage the topic: Firstly the handling of heavy traffic, then splitting responsibilities between coordination and executing forces (aircrafts, ground vehicles) and lastly the introduction of a clear communication standard. In the parking garage the vehicles shall detect and interpret their environment with onboard-sensors to navigate autonomously. More sensors are to be installed inside the park – their purpose is to monitor parking space capacities. Another control system coordinates the gaps between the vehicles. Just like on the street autonomous driving can only work if the car is provided with accurate mapping data. So one of the next steps would be to create HD maps for position tracking.
Now we know what is happening beyond the gates of two driving forces behind autonomous driving. But two more presentations are yet to come – closing up we will show you the presentations of Continental and AUTOSAR, a worldwide development partnership of automotive interested parties working on an open and standardized software architecture for automotive electronic control units.