Structured development and validation of AI solutions

The successful deployment of artificial intelligence does not end with the development of a model – testing in a real-world application context is crucial. AI systems often only demonstrate their true capabilities when deployed under unpredictable conditions, in dynamic environments and amidst complex system interactions. Pure simulations or laboratory environments are therefore insufficient in many cases.

This service provides controlled yet realistic test environments and real-world laboratories in which AI systems, algorithms and data-driven applications can be validated, evaluated and iteratively refined at an early stage. The aim is to reduce technical uncertainties, understand system behaviour under real-world conditions and lay the foundations for safe and scalable deployment in production.

The key added value lies in ensuring robustness, safety and reliability prior to going live. Companies gain early clarity on how their AI solutions behave within the overall system, where the limitations lie and what adjustments are necessary. This reduces operational risks, provides a more solid basis for development decisions and significantly accelerates time to market.

At the same time, real-world laboratories enable continuous, iterative improvement: systems are not merely tested once, but are further developed using real data, real-world environments and actual interactions. This continuous testing and learning process is crucial, particularly in the fields of autonomous systems, robotics, mobility and safety-critical applications.

Examples from the Research Centre for Informatics:

• Testing of mobile robots for inspection and maintenance tasks in real-world operational environments
• Integration and field testing of automated driving and mobility systems (e.g. last-mile transport, logistics and passenger transport solutions)
• Use of reference platforms for sensor, perception and ADAS algorithm testing in the field of autonomous driving
• Use of behaviour trees (e.g. ros_bt_py) for modelling and controlling complex AI-based robotics and automation systems
• ‘Test-before-Invest’ of embedded and edge AI systems, including RISC-V-based energy-efficient hardware solutions
• Simulation and validation of inner-city traffic scenarios in a digital twin of real test routes (including junctions, tunnels and complex traffic conditions)

Who is this service for?

This service is aimed at organisations that not only develop AI systems but also wish to deploy them in real-world or near-production environments. These include industrial companies with automation and production processes, OEMs and suppliers in the mobility and robotics sectors, technology-focused start-ups with hardware or AI products, and research institutions with application-oriented development projects. It is particularly relevant for organisations for whom security, reliability and scalability are crucial, and for whom laboratory testing alone is insufficient.

Cost contribution:

This service is offered as part of the EDIH-AICS 2.0 programme. Costs depend on the scope and effort involved and will be agreed individually with the contact person. Thanks to funding, the service is provided at reduced rates.

Required AI entry level:

This service is suitable for all levels of experience, whether you are a beginner or an expert.

Your contact: