AI platform for autonomous cars reconfigures on the fly
In autonomous driving, data from laser, camera and radar sensors in the car must be reliably and quickly combined and processed. Through this sensor data fusion and intelligent object recognition, the vehicle always has a precise image of the real traffic conditions, can locate itself in this environment and make the right decision in any driving situation on the basis of this information. The data to be processed for the environment recording is so complex that artificial intelligence methods are required to ensure a high level of traffic safety.
Fraunhofer IIS and its partners in the KI-FLEX project are developing a high-performance hardware platform and the associated software framework for this purpose. The algorithms used for sensor signal processing and sensor data fusion are largely based on neural networks and allow the vehicle position and environment to be recorded quickly and accurately.
The significance and usability of individual sensors varies depending on the traffic situation, weather and lighting conditions. In order to do justice to this, the platform is designed as software-programmable and reconfigurable hardware. This means that the algorithms used for sensor evaluation can be exchanged during the journey if conditions change. This allows the car to react flexibly to impairments or even the failure of individual sensors. In addition, the project team will develop suitable methods and tools to ensure the functional safety of the AI algorithms used and their interaction even during reconfiguration while driving. For the efficient execution of all algorithms and reconfigurations, the computing resources of the hardware platform are dynamically allocated according to the load.
The planned platform is a new development in the field of neuromorphic hardware. This functional principle is inspired by the human brain and specially designed and optimized for the efficient use of neural networks. In particular, it takes into account the fact that product cycles in the automotive sector are very long on the one hand, but that AI algorithms are developing rapidly on the other. The development goal of the project is therefore a hardware platform that can be easily and quickly adapted to new software and hardware requirements in the field of machine learning.
Upon request, Fraunhofer IIS provided details: The hardware will be based on a flexibly programmable multi-core deep learning accelerator in the form of a specially developed chip (ASIC). This will consist of a DLI multi-core processor system optimized for image processing and a flexible DLI accelerator core with embedded reconfigurable logic. The integration of a reprogrammable logic core gives the multi-core deep learning accelerator the flexibility it needs to implement new AI algorithms in the future.
Reconfiguration of the platform is purely software-based. Within the framework of the KI-FLEX project, appropriate methods and tools will be developed to ensure the functional security of the AI algorithms in the application.
Fraunhofer IIS will lead the project consortium, which includes a number of research and industry partners such as the lidar specialist Ibeo Automotive Systems GmbH, the chip manufacturer Infineon, the processor IP developer videantis GmbH, the Chair for Robotics, Artificial Intelligence and Real-Time Systems at the Technical University of Munich, the Fraunhofer Institute for Open Communication Systems FOKUS, the Daimler Center for Automotive IT Innovations (DCAITI) at the Technical University of Berlin and the FAU university of Erlangen-Nuremberg.
More information: https://www.iis.fraunhofer.de/en/ff/kom/iot/embedded-ml/neuromorphic/ki-flex.html
Related News:
Xilinx bolsters MPSoC family for more computing power
ZF demos automotive supercomputer, autonomous minibus
Autonomous cars learn to drive with foresight
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
