Edge devices, such as autonomous vehicles, robots, and smart cameras, are typically constrained by their compact size and weight limitations. As a result, they are generally equipped with lightweight chips and multiple sensors to perceive their surroundings quickly. However, current solutions often lack reliability and rely on large, bulky processors that are difficult to integrate into these compact devices, leading to high computational and energy demands.
The vision within the disruptive innovation cycle and AI-driven processes is to enable advanced, reliable, and efficient AI on small-edge devices. This involves significantly reducing implementation costs and eliminating technical and regulatory barriers that have previously hindered widespread adoption. In this project, the Hessian Distr@l program funds the further development and validation of a complete software development kit (SDK) for AI developers to design, optimize, and deploy reliable and energy-efficient multi-task/input AI perception models on small, resource-constrained processors.
Reliable and accelerated AI deployment has the potential to be used in numerous application areas, such as large-scale language models, which have recently become increasingly popular. However, the founding team plans to provide a reliable and fast AI solution, initially focused on the market for camera-based assistants, especially in the areas of autonomous vehicles and intelligent camera systems. The solution addresses design challenges such as limited computing power and energy requirements, making it ideal for autonomous systems. Demand is particularly high in the drone market, where European authorities rely on certified European technology due to strict safety regulations.