PLATON Distributed computing platform for radar-based 3D environment sensing in safe autonomous driving

Goal

In the application area of logistics, Industry 5.0, automated driving, smart construction, smart infrastructure and smart cities, 3D sensors and the efficient and distributed evaluation of the data they generate play a decisive role and the efficient and distributed evaluation of the data obtained by them play a decisive role. The focus is on safety (obstacle detection), mapping and localisation, object recognition, 3D reconstruction, human recognition and similar functions.

In PLATON, research is being conducted into how 3D radar sensor technology can be evaluated in multiple time windows and spatial characteristics using artificial intelligence and how results can be made available for applications. The spectrum of applications ranges from extremely low latency for security applications (obstacle detection) with low spatial resolution to extremely distributed and high-resolution 3D reconstruction. 3D radar technology is particularly data- and energy-saving as well as cheaper than conventional LIDAR methods and promises considerable advantages for diverse areas of application, e.g. in intralogistics and autonomous driving.

More specifically, PLATON will design, implement and evaluate an integrated sensor and processing platform that can dynamically distribute processing loads in networks and operate in an energy-saving manner. This platform enables the simultaneous execution of several AI processes, which can be trained end-to-end in a simulation for virtual commissioning. The user interface is included in the training.

The primary goals of OFFIS in PLATON are in the area of distributed spatial AI. On the one hand, OFFIS is researching the automatic spatial detection of the environment using 3D radar data as well as the robust localisation of the sensor system in this environment. On the other hand, OFFIS focuses on distributed, end-to-end trained, semantic mapping of the environment, which includes object features. This system is being evaluated in the OFFIS laboratories and at project partners.

Persons

External Leader

Matthias Schweiker (PILZ)
Publications
A Coarse-to-Fine Method for Data-Efficient Collaborative Place Recognition

Furuno, Eike and Hein, Andreas and Stratmann, Tim C. and Pfingsthorn, Max; 2024 9th International Conference on Control and Robotics Engineering (ICCRE); 2024

Partners
Pilz GmbH & Co. KG
www.pilz.com
Creonic GmbH
www.creonic.com
let's dev GmbH & Co. KG
www.letsdev.de
Reeb-Engineering GmbH
www.reeb-engineering.de
Fraunhofer-Institut für Zuverlässigkeit und Mikrointegration IZM
www.izm.fraunhofer.de
Fraunhofer-Institut für Hochfrequenzphysik und Radartechnik FHR
www.fhr.fraunhofer.de

Duration

Start: 01.11.2022
End: 31.10.2025

Source of funding