Towards a Quantitative Time Analysis and Decision Support for the Deployment of AI-Algorithms in Distributed Cyber-Physical Production Systems

Modern Cyber-Physical Production Systems get more and more intelligent by higher capacities of the used resources and more resource-efficient AI-algorithms. However, a significant challenge is finding the fitting architecture for hardware and software cost-efficiently and with low effort. Currently, this process consists of trial and error or selecting overpowered hardware resources, which leads to expensive and time-consuming processes in the development. This paper deals with a quantitative benchmark of the timing behavior of selected algorithms for AI and preprocessing on representative hardware platforms in cyber-physical production systems, building on previous approaches that take a model-based view of hardware/software co-design. This approach is a first step away from a purely qualitative system design, towards a quantitative approach.     «

Modern Cyber-Physical Production Systems get more and more intelligent by higher capacities of the used resources and more resource-efficient AI-algorithms. However, a significant challenge is finding the fitting architecture for hardware and software cost-efficiently and with low effort. Currently, this process consists of trial and error or selecting overpowered hardware resources, which leads to expensive and time-consuming processes in the development. This paper deals with a quantitative be...     »