nxtAIM Project – Generative Methods for Perception, Prediction, and Planning. The project kick-off was very successful thanks to the nxtAIM team. All project partners met to jointly initiate the nxtAIM project and discuss the next steps for successful project implementation.
A good atmosphere, a highly motivated project team and an important, current research project for autonomous driving, NXT GEN AI METHODS – nxtAIM, characterised the kick-off on 26/27 February 2024 in Karlsruhe. The various sub-projects were presented over two days and the next steps for the successful implementation of the initiative were discussed, led by project coordinator Jörg Reichardt, Continental and co-coordinator Ulrich Kreßel, Mercedes-Benz. The collaboration between all project members was impressive and promising. The project consortium consists of 20 partners, including car manufacturers, suppliers, technology providers, research institutions, universities and external partners. nxtAIM was initiated and developed from the VDA Leitinitiative autonomous and connected driving and is funded by the Federal Ministry for Economic Affairs and Climate Action. The project will run for 36 months from January 2024 to December 2026.
Automated driving functions are currently still severely limited due to the current system architecture and the discriminative machine learning methods used. nxtAIM introduces generative methods that enable a bidirectional flow of information and will significantly improve the development of autonomous driving functions. Foundation models for driving data will emerge as an outstanding result. The mobility of the future is autonomous, with humans and vehicles continuing to share and interact in the traffic space. Despite many advances, there are still hurdles regarding the scalability, transferability, and traceability of data-driven driving functions.
Generative AI offers an alternative approach and has proven its capabilities in applications such as language models and text-to-image generators. Through generative methods and the development of foundation models, nxtAIM strives for a paradigm shift in system architecture and AI methodology. Current system architectures are based on unidirectional information processing along the chain of effects, from perception and environmental modelling to planning and implementation. Generative methods create a feedback channel and expand information processing bidirectionally. The nxtAIM research is expected to bring improvements in automated driving.
