Together with partners in the KI-FLEX project, the Fraunhofer Institute for Integrated Circuits IIS is developing a platform that uses artificial intelligence methods to help measure vehicle position and determine vehicle environment in the future. The goal is to help ensure that fully automated and autonomous vehicles respond appropriately in every situation.
As part of the project, which runs through August 2022 and is funded by the German Federal Ministry of Education and Research (BMBF), Fraunhofer IIS is leading the development of a software-programmable and reconfigurable hardware platform that processes sensor data with AI-based methods for autonomous driving. The algorithms used for sensor signal processing and sensor data fusion are largely based on neural networks and enable the vehicle’s exact position and environment to be determined. Its instigators say that the project is a key step in the development of technology components that are urgently required to make autonomous driving safe and reliable.
The relevance and usability of individual sensors varies depending on the traffic situation and on the weather and light conditions. To account for this, the platform is being designed as software programmable and reconfigurable hardware, which means that the algorithms used for sensor evaluation can be switched in line with changing driving conditions. This enables the vehicle to respond flexibly if individual sensors are compromised or if they fail.
In addition, the KI-FLEX project team will develop suitable methods and tools for ensuring the functional safety of the AI algorithms used and their interactions, even if the algorithms are reconfigured while the vehicle is on the road. To enable all algorithms and reconfigurations to be executed efficiently, the hardware platform’s computing resources are allocated dynamically according to load.
The planned platform is a new development in the field of neuromorphic hardware, the functionality of which is inspired by the human brain and specially designed and optimized for the efficient use of neural networks.
A key consideration in the project is that while product cycles in the automotive sector are very long, AI algorithms are advancing very rapidly. The project partners are therefore working toward a hardware platform that can be quickly and easily adapted to new software and hardware requirements in the field of machine learning.
To achieve this, they are focusing on using a flexibly programmable multi-core deep learning accelerator in the form of a specially developed chip – ASIC – that helps reduce costs and power consumption compared to conventional multi-purpose processors (CPUs) or graphics processing units (GPUs). On that basis, the project is playing an important role in driving forward both science and the automotive industry in the field of autonomous driving.
