How to eat the autonomous driving verification elephant?

By Paul van Straten

Siemens launches the Simcenter Prescan360 solution to unlock orchestrated cluster-based simulation for large scale virtual verification.

Over the next several decades, the autonomous vehicle market will generate many trillions of dollars of economic activity from new and existing auto makers, shared mobility services, and an increasingly varied auto supply chain. This new autonomous vehicle revolution will also bring big changes to the way the vehicles are designed, manufactured and sold.

Supporting full traceability and verification throughout the entire autonomous vehicle development process, the Siemens Digital Industries Software portfolio supports customers. With an emphasis on an integrated, digital development environment, and advanced chip to full-vehicle simulations, Siemens is helping auto makers and their suppliers create safe, commercially viable autonomous vehicles, on time and on budget.

The highest priority for car companies stepping into the autonomous vehicles world is occupants safety. To conclude on this question, the automotive OEMs will have to check their ADAS/AD systems behavior in a tremendous diversity of situations, with enough simulation fidelity. One talk about tens millions test cases to be checked for highly automated driving technologies to come in the next decade.

Scaling-up virtual verification

In order to support its customers to overcome this unprecedented challenge, Siemens has developed a cluster and cloud-capable, simulation-based verification framework called Simcenter Prescan360.

This solution makes it possible to generate endless numbers of virtual cases out of template scenarios collections, which are the combination of world models with vehicle models connected through sensor models.

Computer-aided engineering teams can use Simcenter Prescan360 and Simcenter Amesim to model the autonomous system or even assess the complete vehicle. As a part of Simcenter Prescan360, Simcenter Prescan is the core simulation engine for environments, scenarios, sensors, AI and controllers, and its representations contain the reflective properties for radar, camera, lidar and ultrasonic sensors. The sensor models can therefore be scaled from basic ground truth information up to full-wave propagation modeling for detailed sensor evaluations.

Scenarios can be created manually, in a scripted way, or taken from databases such as GIDAS and CIDAS (German and Chinese accident databases, respectively). Also, many OEMs and Tier 1s record traffic data and generate OpenScenario format descriptions, which Simcenter Prescan360 can import.

HEEDS, Simcenter Prescan360’s orchestrator aims to check requirements through model-in-the-loop or software-in-the-loop simulation. HEEDS has evolved from its previous design toward massive simulation orchestration, specialized for ADAS and autonomous driving virtual assessments. It offers mature technologies for process automation, execution distribution on any targets, and high-level results dashboards, useful for extensive analysis of numerous results.

HEEDS is also able to efficiently explore a scenario’s parametric space to capture valuable simulation results, giving precious insight about failure modes of ADAS/AD systems under development.

Simcenter Prescan360 is also very open to third-party tools and scripts to customize the processes for all companies and simulation teams. Simcenter Prescan360 is a cluster and cloud capable simulation solution, with local machine orchestration still available for short assessments and simulation debug. Simcenter Prescan360 can be deployed and used with on-premises clusters, distributed computing or special machines with some deployment tailoring Siemens provides. It can alternatively be deployed and used out of the box on a Microsoft-managed Azure cloud, which makes a lot of sense as a first cloud simulation-as-a-service experience, by outsourcing execution resources scaling and maintenance, while refining the execution platform requirements for later on-premises investment. Siemens plans to address other cloud providers in the future with Simcenter Prescan360.

If more physical properties are needed for the vehicles, Simcenter Amesim is the tool of choice for vehicle dynamics and powertrain modeling. It offers the possibility of modeling car physics and controls, using various fidelity levels that best match any simulation activity.

Simcenter Prescan360 for ADAS/AD was just born and made to confirm autonomous safety and performance, supporting Siemens’ customers to introduce safe, reliable and comfortable self-driving cars.

Read more about Simcenter Prescan360 in this white paper.

Learn more about Simcenter Prescan here.

Read more about HEEDS here.

This article first appeared on the Siemens Digital Industries Software blog at