TUM researchers develop new technology to improve road safety for autonomous driving

According to foreign media reports, in the Providentia++ project, researchers from the Technical University of Munich (TUM) and industry partners have developed new technologies to supplement the vehicle's perspective based on on-board sensor input and a bird's-eye view of traffic conditions, thereby improving road safety, including autonomous driving.

“Autonomous vehicles must be safe not only at low speeds, but also at high speeds,” said Jörg Schrepfer, head of advanced driving research at Valeo Germany. “But when objects fall from a truck or in other situations, the ‘egocentric’ perspective of an autonomous vehicle often fails to detect danger in time, making it difficult to perform smooth evasive action.”

Image source: Technical University of Munich

Researchers in the Providentia++ project have developed a system that transmits an additional view of the traffic situation into the vehicle. "By using sensors on the elevated sign bridges and masts, we have created a reliable real-time digital twin of the test route that can be operated 24/7," said Professor Alois Knoll, project manager at TUM. "With this system, we can supplement the vehicle's view with an external perspective (bird's eye view) and incorporate the behavior of other road users into decision making."

Transferring the digital twin to the car: minimizing time lags

Transmitting the digital twin to the car is no easy task, as it requires knowing the exact position of the vehicle to which the sensor station information is transmitted. To this end, project partner Valeo uses an IMU-GNSS system (Inertial Measurement Unit – Global Navigation Satellite System) consisting of a measurement unit, a satellite navigation system and a real-time motion kit.

"In this way, we can create a coordinate system that is accurate to the centimeter in real time," says Valeo expert Jörg Schrepfer. To synchronize the information from the vehicle and the measuring station, the researchers use the UTC standard, which provides a uniform basis for coordinating time. Ideally, the digital map would be superimposed on the car's view as a second layer.

But time lags (latencies) in the entire system cannot be completely avoided. There is still a long time difference from the physical detection and data processing of the sensor to the radio transmission to the vehicle. The data needs to be packaged, encoded and transmitted, and then decoded in the car. Other issues also affect the length of time, such as the distance of the vehicle from the transmission tower on the test route and the traffic volume of the data transmission network. In a recent demonstration run, Valeo used the LTE (4G) wireless standard, which produced delays of 100 to 400 milliseconds. Schrepfer explained: "These delays can never be completely eliminated. However, smart algorithms can help. When we have full coverage of 5G or 6G telecommunications standards, the time delays will be shorter."

Prototypes ready for real-time digital twins

The Providentia++ research project creates the conditions for using this data in vehicles. The project goal is to create a scalable and highly available digital twin of traffic situations with real-time capabilities. To this end, the team built a 3.5-kilometer test route in Garching, outside Munich, including seven sensor stations. The prototype was developed to carry out a series of practical operations when needed:

• The researchers will use a decentralized digital twin that will allow the test route to be scaled up or extended to any desired length.

  
  

• To handle data volumes of several gigabytes per second, researchers have created a data processing concept that optimizes the load distribution across multiple CPUs and graphics cards (GPUs).

"The digital twin is ready for the project development phase," said Professor Alois Knoll, head of the alliance at TUM. "The concept will enable reliable operation 24/7, not only on motorways but also on secondary roads and near intersections."