Porsche Engineering Turns to AI to Improve AV Simulation

AVEAS research project includes 20 partners to generate more data aiding simulation of critical traffic scenarios.

December 6, 2023

Bill Visnic

The AVEAS research project is striving to expand the data available to create simulations of critical traffic situations. (Porsche)

Porsche Engineering has teamed with a variety of partners for a new research project seeking to enhance the data available for simulating automated driving systems’ reaction in critical situations. The company said its AVEAS project –— the acronym comes from German words for “collecting, analyzing and simulating traffic situations relevant to validation” — is working to develop “automated detection of critical traffic situations from sensor data using AI and storing the situations in a database. The route models and traffic situations generated in this way are also varied in order to generate more test cases for virtual validation.”

Artificial intelligence is being trained to evaluate sensor data from real-world driving and extend that information into simulation that generates more-critical situations. (Porsche)

It's widely conceded that simulation is essential to comprehensively validate an automated-driving system or sensors’ ability to properly and safety perform in the almost infinite number of driving scenarios that might occur on all the world’s roads. But one problem with simulation, Porsche Engineering said, is the difficulty in collecting “critical traffic situation” data from real-world journeys on which a multitude of variations then can be safely enacted in simulation. “After all,” the company said in a release, “in normal traffic situations, marginal situations rarely occur.”

The company said the project’s goal is to automatically evaluate real-world test drives. That baseline data is used to create critical traffic situations as simulated scenarios. For example, engineers can specifically increase criticality by reducing the distance between vehicles.

“We are building a complete catalog of critical scenarios that enable us to validate driver assistance systems and functions for highly automated driving,” Joachim Schaper, head of AI and Big Data at Porsche Engineering and Tille Karoline Rupp, responsible for Simulation at Porsche Engineering, said in a statement.

Porsche goes to Jupiter

The “X” in the lower right corner indicates that virtual test environment has determined the automated-driving function needs to be improved. (Porsche)

Porsche Engineering said it is contributing several key components to the research effort. A JUPITER (Joint User Personalized Integrated Testing and Engineering Resource) test vehicle is being used for the real-world test drives that collect vital sensor data. The JUPITER vehicle equipped with cameras, radar, and lidar sensors; data from those perception components is relayed to the cloud. Porsche Engineering also manages evaluation of the data collection, as algorithms record the road itself and the position of other road users and their behavior. Porsche Engineering said the machine-learning methods used “are constantly being refined.”

The AI-supplemented system applies an added level of “prediction” that a conventional simulation may not be equipped to offer. “We are currently developing a method that continues to recognize road users, even when that category of road user has not been seen for a long time – for example because they have been hidden behind a truck,” said Leon Eisemann, a doctoral candidate and specialist in image recognition at Porsche Engineering.

Tille Karoline Rupp, responsible for simulation at Porsche Engineering. (Porsche)

The recorded traffic events are stored in standardized file formats such as ASAM OpenDRIVE (logical description of the road network) or ASAM OpenLABEL (objects and their dynamics). “AVEAS can therefore also provide input for other projects, such as route modeling,” the company said. In a second step, algorithms select the critical traffic situations by focusing on, for example, short clearances or unusually intense deceleration forces.

‘Trajectory description’

For one example, validating the reaction of adaptive cruise control to avoid hitting the tail end of a traffic jam would be a relevant critical scenario. Environmental factors can also trigger critical scenarios: when a vehicle approaches the exit from a tunnel, for instance, the sudden glare may dazzle onboard cameras.

Porsche said AVEAS’ selection algorithm also highlights those kinds of traffic situations so that they can be used to safeguard driving functions. The ideal is for an automated-driving vehicle to react as proactively as a human driver, for example by reducing speed or prioritizing other sensor inputs. First, the simulatable driving scenarios consist of position data gleaned from road users over time; experts call this a ‘trajectory-based’ description.

The virtual test drives take place in the internally developed simulation environment known as PEVATeC SimFramework (Porsche Engineering Virtual ADAS Testing Center Simulation Framework). The real journey can be reconstructed — simulated — and then played through after specific modifications have been made, all within the digital world. “In what is known as scenario sampling, the real critical situations are systematically modified, which artificially extends the virtual validation test space,” explained Rupp.

For real-world and virtual test drives to be consistent, a considerable degree of “experience” is required — much of which must be gained in the analog world. “What is needed is a deep understanding of how real technology and simulations are related,” Schaper stressed. A virtual The project, which was launched in December 2021, has already delivered its first results, Porsche said. “There are many links in the process chain, test drives are underway, and some patents have already been filed,” says Michael Strobelt, who coordinates Porsche Engineering’s participation in AVEAS.