Departement Human-Cyber-Physical Systems (HCPS)

cognitive system engineering

To ensure that digitalization is aligned with people and benefits them, a holistic, structured approach is required. The upper part of the figure illustrates the simplified, human-centric "Human-CoMo" digitalization cycle, which can be used, for example, to develop decision-support systems. The various roles played by humans involve distinct areas of expertise and tasks that need to be integrated.

The lower part of the figure shows an excerpt from a hierarchical cognitive map representation of expert knowledge — using a galvanic coating process as an example — as a partial result of a knowledge elicitation process. To the right is a corresponding data driven AI model built upon this knowledge model. The underlying data was compiled in collaboration with the B+T Group.

The HCPS department has expertise in the areas of “cognitive systems engineering” and “AI modeling”. They form the basis for making empirical knowledge usable for various digitalization solutions (for example implementation of implicit knowledge in chatbots). Further on the development of AI-based decision support systems and research on “human-machine teaming”.

own references:

• Bocklisch & Lampke (2024) "Human-CoMo"
• Bocklisch (2025)

decision support

Using a mock-up example of multi-criteria parameter optimization in electroplating processes, the video demonstrates how human-centered digitalization in production can be successfully achieved through assistance solutions.

It addresses requirements that arise when deciding between input parameter sets or alternative electrolytes or manufacturing processes, using galvanic zinc-nickel coating as an example.

The focus is on the target criteria of "coating thickness," "nickel content," "corrosion protection," and "surface roughness"—examples of key coating characteristics that often conflict with one another during practice. The assistance system enables the optimization of each of these criteria — either individually or on a multi-criteria basis — thereby allowing for the realization of various target profiles.

Based on existing data from coating experiments conducted in collaboration with the B+T Group, the AI ​​analyzes which combinations of process parameters align with the selected objectives. Subsequently, specific recommendations for parameter settings are derived and presented transparently to the human decision-maker. It is possible to quantify the impact of variations in input parameters on individual target variables, thereby identifying cause-and-effect relationships.

own references:

• ZVO Report 02/2026
• Bocklisch et al., 2024
• Bocklisch et al., accepted

human-machine-teaming

The forms of collaboration between humans and machines are undergoing constant and rapid change. Alongside assistance and automation solutions, human-machine teaming is emerging as a new, highly innovative approach within research and development. Teaming capable systems aim to enable humans and the cyber team partners embedded within technical systems to collaborate much like a well functioning team of colleagues. This requires a stable partnership between high-performance technology and skilled humans—a collaboration that will unlock entirely new possibilities regarding performance capabilities and sustainable work practices.

For a system to be capable of teaming, it must meet among others the following criteria:

• complementarity: Humans and cyber team partners contribute complementary functions to the work process to accomplish tasks. Enhanced performance arises from combining their respective strengths.
• shared knowledge and common Goals Both sides must understand the context of the collaboration and work towards shared goals. In addition to their respective domains of knowledge, there is a shared knowledge base grounded in human expertise and its cognitive structures. Its implementation within the cyber team partner is transparent and both comprehensible and traceable for the human.
• constrained autonomy As in a human team, the partners are interconnected and bound to one another. Each partner has scope for action commensurate with their tasks and strengths— ,overall the autonomy of the team partners is limited. The cyber team partner can be overridden by the human, ensuring that decision-making authority always rests with the human team partner.
• trust in Automation Justified human trust in the cyber team partner emerges from the successful realization of the criteria "complementarity," "shared knowledge/goals," and "bound autonomy" within actual interaction experiences in production-related use cases.

own references:

• Bocklisch & Huchler (2023)
• Bocklisch (2025)

human centered modeling and artificial intelligence

https://www.mec.ed.tum.de/utg/dfg-schwerpunktprogramm-2476/teilprojekte/projekt-4/

Complex manufacturing processes or process chains can be partially modeled or simulated using data-driven AI methods. The resulting models should be transparent and easily understandable to humans. In this way, model results can, for example, provide targeted support for the industrial manufacturing of components and products with defined property profiles. Model-based description is essential for addressing challenges such as the optimization of individual manufacturing processes or global optimization across the entire process chain, as well as for managing inherent process uncertainties and heterogeneous data.

The figure illustrates a specific example of a process chain comprising (1) thermal spraying, (2) turning, and (3) diamond burnishing, designed to produce a functional surface with a graded increase in hardness near the surface. Thermal spraying (primary shaping) is used to apply manganese hard steel surfaces to rotationally symmetric components, while turning and diamond burnishing are employed to adjust core and surface properties. In this example, the target parameters of the process chain are to be influenced as follows: (1) maximize surface hardness, (2) minimize surface roughness, (3) maximize the oxide content of the coating system, and (4) maximize the energy efficiency of the surface generation process.

Integrated sensor systems capture manufacturing process data, which is then used for modeling. The methodological approaches employed include statistical modeling and multidimensional, pattern-based description using grey-box AI algorithms (fuzzy pattern classification). These methods enable the quantification of various types of uncertainty as well as the forward and backward coupling of the process chain.

Model based decision proposals assist in research and development by addressing even highly complex issues that go beyond knowledge based or heuristic decision making, while remaining closely aligned with the human perspective.

cognitive ergonomics

https://www.tu-chemnitz.de/mb/WOT/abteilungen/hcps/forschung/projekte/komti/demonstrator/

The design of complex human-AI-technology systems must be guided by human-centered criteria to ensure that the human operator can perform their tasks optimally without being subjected to undue strain. Incorporating insights and methods from cognitive psychology can foster new qualities of interaction between humans and (semi-)autonomous technical or digital systems, while also increasing the acceptance of the technology.

The research project KoMTI "Kognitionsbasierte Mensch-Technik-Interaktion" investigated a holistic approach to the human-centered design of technology, guided by the principle of helpful technology that supports human self-determination rather than controlling or restricting it. The project resulted in a modular methodological toolkit and a procedural strategy for developing and implementing human-oriented interaction interfaces within complex human-technology systems.

Technologies and human-centered measures—such as those promoting sustainable, resource-efficient practices in day-to-day production—are successful only if people understand, accept, and implement them. To this end, the HCPS department employs and adapts a variety of methods. These include observation and survey techniques, interviews, psychological experiments, and user studies, as well as cognitive task analyses, procedures for assessing (team) situation awareness and mental workload, and eye-tracking-based measurements of eye movements and pupil dilation./p>

The goal of the research project REDUCE was to reduce energy consumption in industrial production. To this end, a transdisciplinary approach combined technical possibilities with a human-centered perspective to develop a human-machine interaction (HMI) demonstrator that visualizes energy consumption and provides actionable recommendations for the user. This demonstrator was subsequently evaluated under real-world operating conditions in terms of efficiency, user experience, acceptance, and individual preferences.

own references:

• Link zum KoMTI-Demonstrator
• Abschlussbericht des Projektes "REDUCE"