Architecting a Centralized, Scalable and Modular Platform for Failsafe Autonomous Driving
By Matthias Schulze, Vice President Autonomous Driving Technology Management, Visteon
Matthias Schulze, Vice President Autonomous Driving Technology Management, Visteon
Visteon has developed a unique centralized computing approach based on an open platform that allows automakers, developers and technology suppliers to collaborate on innovations addressing the new demands of autonomous driving. DriveCore™’s modular, flexible approach is designed to meet all levels of automation from Level 2-5 – providing a tremendous asset compared to currently available offerings.
Visteon’s DriveCore™ is a newmodular and flexible approach for a centralized ADAS and autonomous driving domain controller that provides an open computation platform. Incorporating unique DriveCoreRuntime™ middleware with open application programming interfaces, Visteon’splatform allows easy assimilationof software components from automakers, suppliers and third parties. Algorithm development for DriveCore™ is supported by DriveCoreStudio™, Visteon’s own software development kit (SDK), whichallows developers to run, modify, optimize and benchmark ADAS and autonomous driving algorithms in a safe PC environment with real life and synthetic sensor data. As a result, theDriveCore™ platform accelerates the development of driver assistance systems and systems for autonomous driving ranging from Level 2-5. DriveCore Studio is also offering a cloud extension, which allows several developers to develop algorithms on the same DriveCore Compute unit, running several Studio SDKs in parallel in the cloud to test several algorithms in parallel and to do over-the-air updates of Visteon´s DriveCore Compute units.
DriveCoreCompute™ - centralized computing for maximum flexibility
Autonomous vehicle development is prompting new open architectures, driven by increased collaboration with multiple developers to meet automakers’ need to differentiate. Such software flexibility is not possible with today’s distributed electronic control units (ECUs) but is the hallmark benefit of centralized computing units.Centralized computing – using domain controllers – provides unprecedented flexibility, power and speed for handling the vast amount of sensors required for failsafe autonomous driving.
Visteon’s DriveCore Compute™ hardware offers maximum flexibility through its unique modular design that on the A-sample level consists of a baseboard and three types of daughter boards:
• A computation carrier, that can incorporate any system-on-chip (SoC) and is key to Visteon’s flexibility in choosing chipset vendor
• A safety carrier with an ASIL D-certified chipset that provides the ASIL D compliance needed for highly automated and autonomous driving functions
• An Ethernet extension board that allows a magnitude of sensors to connect with the domain controller
Up to eight daughter boards can be stacked on one baseboard through Visteon’s unique connector system. This makes DriveCore Compute™ scalable to enable all levels of automation – ranging from driver assistance to fully autonomous driving – in an efficient and cost-effective way.
The modular approach with multiple boards of DriveCore Compute on A-sample level enables rapid prototyping of customer specific solutions. B and C-samples will be one PCB units with the form factor of the future production ECU.
DriveCoreRuntime™ – unique middleware for high frequency and low latency data exchange
DriveCore Compute™ comes with DriveCore Runtime™ - Visteon’s own middleware with a unique shared memory concept. Runtime has been developed to accommodate the huge amount of data that is present when in anautonomous driving system the data of a multitude of high resolution sensors is fused on raw data level in the domain controller as it is intended by Visteon for our autonomousdriving software stack. This is particularly important when artificial intelligence (AI) is applied for decision making. Only DriveCoreRuntime™ is offering the high-frequency, low-latency communication layer that enables this task.
By offering the same comfortable interfaces as ROS (Robot Operating Systems), but being automotive grade and ASIL B-certified, Runtime also accelerates algorithm development considerably. In addition, Runtimeincorporates open Application Programing Interfaces (APIs) which allow for easy integration of third party algorithms, making DriveCore™ a truly open computing platform.
DriveCoreStudio™ – ADAS software development kit with unique features
WithDriveCoreStudio™, the DriveCore™ platform provides a unique set of tools for developing, verifying and optimizing software in an open platform environment. The DriveCoreStudio™ software development kit allows developers to test, modify, visualize and evaluate ADAS algorithms in a safe PC environment; algorithms can even be run in parallel for comparison and benchmarking.
For this, Studio allows a feed of real life or synthetic sensor data and provides interfaces to simulation tools such as VTD, Car Maker or Oktal. In addition, profiling capabilities are provided that help algorithm developers to optimize their algorithms with regard to CPU, GPU and memory usage.
Algorithms developed with DriveCore Studio™ can be uploaded and run directly on the DriveCore Compute™ ADAS domain controller.
Besides the capabilities to test modify, visualize and profile ADAS and autonomous driving algorithms, DriveCore Studio can also be used as data recording tool and, with its cloud extension, it even allows several software developers to work on one DriveCore Compute unit only. This is of particular importance in early project phases, where only a limited number of prototype ECUs is available. The cloud extensions do also allow to download test data from repositories in the cloud or from 3rd party data bases and to run several DriveCore Studio SDKs in parallel in the cloud to test and validate a large number of algorithms in parallel. Also over-the-air update of DriveCore Compute ECUs is enabled through the cloud extensions of DriveCore Studio.
ADAS and Autonomous Driving algorithm development at Visteon
Visteon’s in-house development of algorithms for driver assistance systems and autonomous driving draws heavily on the application of state-of-the-art AI technology, which is complemented by deterministic methods - where this is needed to fulfil the stringent functional safety requirements for autonomous driving systems.
Visteon has developed all software components needed for a Level 3 Highway Pilot and Traffic Jam Assistance system that has left prototype status. Work is focussing now on the extension of the algorithm basis towards automated parking functions. Full Level 4 functions – first for motorway applications and later for urban driving – are planned to follow.
Development centers all over the world with local test vehicles guarantee that Visteon’s ADAS systems work on all places, where automated driving is expected to play a significant role in road traffic in the foreseeable future.
Up to eight daughter boards can be stacked on one baseboard through Visteon’s unique connector system. This makes DriveCore Compute™ scalable to enable all levels of automation – ranging from driver assistance to fully autonomous driving – in an efficient and cost-effective way.
The modular approach with multiple boards of DriveCore Compute on A-sample level enables rapid prototyping of customer specific solutions. B and C-samples will be one PCB units with the form factor of the future production ECU.
DriveCoreRuntime™ – unique middleware for high frequency and low latency data exchange
DriveCore Compute™ comes with DriveCore Runtime™ - Visteon’s own middleware with a unique shared memory concept. Runtime has been developed to accommodate the huge amount of data that is present when in anautonomous driving system the data of a multitude of high resolution sensors is fused on raw data level in the domain controller as it is intended by Visteon for our autonomousdriving software stack. This is particularly important when artificial intelligence (AI) is applied for decision making. Only DriveCoreRuntime™ is offering the high-frequency, low-latency communication layer that enables this task.
By offering the same comfortable interfaces as ROS (Robot Operating Systems), but being automotive grade and ASIL B-certified, Runtime also accelerates algorithm development considerably. In addition, Runtimeincorporates open Application Programing Interfaces (APIs) which allow for easy integration of third party algorithms, making DriveCore™ a truly open computing platform.
DriveCoreStudio™ – ADAS software development kit with unique features
WithDriveCoreStudio™, the DriveCore™ platform provides a unique set of tools for developing, verifying and optimizing software in an open platform environment. The DriveCoreStudio™ software development kit allows developers to test, modify, visualize and evaluate ADAS algorithms in a safe PC environment; algorithms can even be run in parallel for comparison and benchmarking.
For this, Studio allows a feed of real life or synthetic sensor data and provides interfaces to simulation tools such as VTD, Car Maker or Oktal. In addition, profiling capabilities are provided that help algorithm developers to optimize their algorithms with regard to CPU, GPU and memory usage.
Algorithms developed with DriveCore Studio™ can be uploaded and run directly on the DriveCore Compute™ ADAS domain controller.
Besides the capabilities to test modify, visualize and profile ADAS and autonomous driving algorithms, DriveCore Studio can also be used as data recording tool and, with its cloud extension, it even allows several software developers to work on one DriveCore Compute unit only. This is of particular importance in early project phases, where only a limited number of prototype ECUs is available. The cloud extensions do also allow to download test data from repositories in the cloud or from 3rd party data bases and to run several DriveCore Studio SDKs in parallel in the cloud to test and validate a large number of algorithms in parallel. Also over-the-air update of DriveCore Compute ECUs is enabled through the cloud extensions of DriveCore Studio.
ADAS and Autonomous Driving algorithm development at Visteon
Visteon’s in-house development of algorithms for driver assistance systems and autonomous driving draws heavily on the application of state-of-the-art AI technology, which is complemented by deterministic methods - where this is needed to fulfil the stringent functional safety requirements for autonomous driving systems.
Visteon has developed all software components needed for a Level 3 Highway Pilot and Traffic Jam Assistance system that has left prototype status. Work is focussing now on the extension of the algorithm basis towards automated parking functions. Full Level 4 functions – first for motorway applications and later for urban driving – are planned to follow.
Development centers all over the world with local test vehicles guarantee that Visteon’s ADAS systems work on all places, where automated driving is expected to play a significant role in road traffic in the foreseeable future.
Read Also
Technology Interwoven with the Fabric of Cars
James Seevers, CIO & GM, Toyoda Gosei
A Global Understanding Imperative in the Automotive World
Dennis Hodges, CIO, Inteva Products
The Future of Autonomous Transportation Requires AI Supercomputing in the Car and in the Cloud
Serkan Arslan, Director of Automotive, NVIDIA EU
How Co-Simulation Accelerates the Future of Autonomous Driving
Dr. Thomas Benz, Solution Director Automotive, PTV Group
