The quality assurance strategy that is nowadays used for mass production, based on fixed metrological configurations and rigid centralised production control, is unable to cope with the inspection flexibility needed among automated small series production (SSP). The major challenge faced by a quality assurance system applied to small series production facilities is to guarantee the needed quality level already at the first run (“first time right on time”). In this sense, the quality assurance system has to adapt itself constantly to the new production conditions and support continuous process improvement. The cognitive metrology project aims to develop methods, technologies and services for an efficient quality assurance and metrology application for flexible small series production, able to adapt themselves to changing requirements based on cognition regarding product and process. The project is based on two applications: a PCB production line in Brazil and an inspection machine for transparent freeform parts in Germany. The project addresses two main research areas: quality prediction/ planning and as well as measurement and control systems. Three key elements were developed within the scope of the project. These include a holistic approach to flexible control systems, thus creating the necessary environment for SSP inspection, expert systems for the extraction of information for the control system and the quality planning steps and lastly the development of a flexible metrology infrastructure based on different sensor technologies and sensor data fusion. A flexible and generic service-oriented (e.g. multiagent) control structure for implementing the concept of cognitive production metrology was implemented and validated in both use cases and is adaptable to other small series production layouts. Two different approaches to the design of agent based control systems were discussed for the two use cases. The debate revealed numerous influencing factors that need to be considered when designing MAS. Accordingly, a guideline was developed to help production engineers in the design specifications of agent systems. Expert systems of non-probabilistic (e.g. rule-based) and probabilistic (e.g. Bayesian networks) nature were applied for various tasks in the project. Most notably, a Bayesian network was designed for the determination of probable causes for defects in the PCB scenario. The network was tested successful and helped to reduce the number of possible causes in a complex interlinked scenario with 29 defects and 69 possible causes. A generic framework for sensor data fusion was established and applied in two different scenarios. One utilizes a number of raw images obtained with different illumination setups to create a single data source for the quality evaluation of headlights. The other fused extracted features from 2D images and a 3D topology to determine a 3D-structure tensor to analyse the fibre orientation in CFRP prepregs. The obtained results enhance the process efficiency of the flexible small series production considerably by using adequate metrology and quality assurance solutions supporting selfoptimising strategies, aiming at e.g. a faster and easier production and inspection system setup time, a reduced measuring time and complexity, supporting an improved number of product variants and autonomous inspection planning. The results of this project form major steps towards the vision of cognitive production metrology. To fully achieve this goal further research in the field of flexible control system and algorithm design expert systems and autonomous configuration of inspection systems based on fully parametric setups is required. These further developments will contribute directly to reduce the complexity of pilot production series, to speed up the production start time and to assure a maximum quality level for the process and product.