Our vision
„The critical path to introduce autonomous vehicles to the market is not in the technology, but in developing a metric that enables an approval!“
(Prof. Dr. Winner: 6. FAS-Tagung München, November 29th 2013)
At the AVL Mobility and Sensor Center Roding, we meet this challenge and realize:
- Sensor validation in reproducible weather situations
- High-precision calibration of reference systems
- Collection of ground truth reference data on behalf of customers
- The development of innovative mobility concepts for rural areas
Sensor Test Center
In the Roding sensor test center, we are creating a test environment with our cooperation partners from industry and science which is to date unique. In addition to sensory systems, the perception algorithms of commercial manufacturers are also independently tested here.
As the first industrial homologation center for AD systems in Bavaria, we are making a decisive contribution to the certification of future automated vehicles. Thus, we create a basis for new test methods and standards and secure technological progress as well as qualified jobs in the region.
Sensor validation in reproducible weather situations
Indoor AD test hall with pilot character
The Roding sensor test center is one of the first indoor test halls in Germany. With this test center, we meet the industry’s need for independently conducted sensor tests for automated vehicles. The tests consider different weather and traffic conditions. For homologation, these also cover the existing requirements for repeatability, traceability, and documentation.
AD system test under reproducible weather conditions
The “System under Test” (SUT) is subjected to a complete AD system test and tested with the following parameters:
- critical weather conditions such as fog, rain, and back light
- various traffic situations such as spray
- reference vehicle (ego vehicle)
Reliable ADAS / AD Validation Toolchain
After completion of the tests, the collected data from the different sources and measurements are summarized, processed in the data center and evaluated in an automated and real-time toolchain. In addition, real driving tests at higher speeds can be carried out on an outdoor test track using a mobile rain system.
The combination of virtual and real driving tests enables resilient, reproducible and efficiently implementable tests for autonomous driving vehicles – and for our customers reliable results for their further design and homologation process in the new autonomous era.
As one of the world’s leading providers of simulation and test systems for the automotive industry, AVL sets methodological standards to meet the challenges of validating autonomous driving vehicles. We develop our innovative test solutions in the Sensor Test Center Roding in cooperation with leading universities, research institutes as well as our customers.
Sensor Calibration
Precise sensor perception
To achieve accurate measurement results, perfectly matched coordinate systems of the sensor systems to be compared are a decisive factor. Even minimal deviations lead to significant perception errors and falsify the ground truth.
High-precision calibration in 3 steps
The sensor calibration system enables efficient and accurate calibration of the sensor systems to be compared by combining:
- Intrinsic camera calibration: correction of the geometric distortion of all cameras
- Extrinsic sensor calibration: Alignment of the coordinate systems of the reference sensors to each other
- Extrinsic DGT-HW to test vehicle calibration: Alignment of the coordinate system of the reference system to the vehicle coordinate system
AVL Dynamic Ground Truth System
Accurate reflection of reality – ground truth
The objective comparison of sensor signals with an absolute environmental reference (ground truth data) shows how accurately a sensor system captures its environment – and at the same time enables cost-effective development and validation of sensor technology at all levels of the development process. Certification authorities also need to be able to independently evaluate sensor systems for road approval of future autonomous driving vehicles. A high-precision sensor reference system is the basic prerequisite for this.
The future standard
The AVL Dynamic Ground Truth System (DGT) captures an exact 360° field of view of the vehicle environment for statistical comparison and validation of the ADAS/AD system under test (SUT). The captured image is composed of time-synchronized lidar, camera, and high-accuracy GNSS data to enable precise comparison of the DGT and SUT systems.
Using our powerful Perception software, the independent reference image of the vehicle environment is generated based on the recorded raw data. The result is an OSI data stream containing information about recorded objects such as cars, pedestrians, motorcycles, and traffic signs. These are used for analysis and validation against the SUT – taking us one step further towards Data Driven Development.
More informationen at:
https://www.avl.com/-/dynamic-ground-truth
https://www.avl.com/-/ground-truth-measurement-for-data-driven-development
Digital twin of the AutBus route in Neubäu am See (Bavaria)
To ensure efficient software function development, the precise mapping of the autonomous driving route in a simulation environment is essential. With the help of this digital twin, software functions such as the lidar-based localisation, which is calculated redundantly to the GNSS-based localisation, can be tested in advance.
In contrast to testing such functions on the basis of real data, the so-called ground truth (i.e. the real positions, orientations and dimensions) of all objects is available in the simulation. Structured and automated parameter variations (e.g. changed parking situations of vehicles at the roadside or changed seasons with regard to the canopy of trees along the route) can also be carried out much more easily, quickly and cost-efficiently than with the use of real driving data.
The digital twin for the route of the AutBus project in Neubäu am See was implemented with a combination of esmini (parameter variation), AVL VSM (vehicle dynamics) and Unreal Engine (graphic representation).
Digital twin of Euro NCAP – scenarios in varying weather conditions
Currently, ADAS safety systems such as Emergency Braking Assist are only tested in good weather conditions (and in some cases in the dark) as part of the Euro NCAP AEB VRU test procedure. However, a significant proportion of serious and fatal accidents, particularly involving vulnerable road users, occur in the context of bad weather and/or poor visibility.
The AVL Mobility and Sensor Center in Roding offers the possibility to reproducibly test and evaluate the performance of active safety systems even under adverse weather conditions such as rain, fog and varying lighting conditions.
For further parameter variation, the environment of the indoor test facility as well as the depicted weather situations are completely mapped as a digital twin. In this way, parameters such as the pedestrian’s clothing in combination with the weather can be systematically altered in the simulation. The results of the simulation are confirmed and their quality improved by real tests carried out at defined support points.
Milestones
