Disclaimer
A note in advance: this article is the preprint version of the article for the journal “Autismus” of the Bundesverband Autismus Deutschland e. V., translated into English. In addition, below you will find some CC-BY-4.0-licensed material for information and further use. On the one hand, these are the slides and handout from the project’s closing event at the SRH Berufsbildungswerk Dresden on 24 March 2026. On the other hand, these are the autism personas mentioned in the text, in both German and English.
People on the autism spectrum face a wide range of everyday challenges that are often barely visible to outsiders, yet nonetheless lead to considerable strain. This effect is particularly evident in areas of existential importance, such as professional development in adulthood. The employment situation is a clear indicator of an existing imbalance: robust figures are scarce, but estimates suggest that around a third to a half of people on the autism spectrum are unemployed, and only a small proportion – around 5 to 10 per cent – are employed in the general labour market (Bader et al., 2018). This imbalance is also apparent in workshops for people with disabilities, where the work carried out is often well below the person’s actual qualifications. The underlying cause lies primarily in the working conditions of the general labour market, which frequently fail to accommodate the personal challenges faced by autistic people.
This is the starting point for the AutARK research project at the TUD Dresden University of Technology. The project name is a German acronym for “automatic adaptation of overstimulating contexts”, although over the course of the project it became clear that technology could address not only sensory overload but also a range of further topics. The project was funded by the Federal Ministry of Labour and Social Affairs from compensatory levy funds. It ran from 2023 to early 2026 (Technische Universität Dresden, 2026).
A human-centred development approach was fundamental to the project: the aim was to develop technologies that address the specific challenges faced by autistic people while adequately reflecting the individuality of the spectrum. A further key consideration was that personal support services, such as workplace assistance, are constrained by limited and costly staffing capacity. Technical solutions, by contrast, are more flexible, strengthen a sense of self-determination and self-efficacy, and can already be used during school, training or study, and ideally carried over into the workplace later on.
Another hallmark of the human-centred approach was that the target group was actively involved throughout the entire development process: rather than talking about them, challenges were identified and solutions developed together with them, for example through interviews and observational studies. This involvement increases both the acceptance and the accuracy of the resulting solutions.
Over the course of the project, numerous concepts were developed and tested in studies with users. The resulting solutions fall into three thematic areas: communication and social interaction, sensory perception and processing, and time and task management. More on this shortly. The response at the closing event at the SRH Berufsbildungswerk Dresden was particularly encouraging: more than 120 participants from vocational rehabilitation, training, and business showed keen interest – a clear signal for us as a research team to continue pursuing the concepts developed and to bring them into practice.
The starting point: listening and understanding
It is a fair question why a team of (media) computer scientists should develop assistive systems for such a sensitive target group without being autistic themselves. To answer this, it is worth looking at a particular area of applied computer science: human–computer interaction (HCI). Within HCI, detailed and methodologically rigorous procedures have become established, especially regarding digital accessibility, ensuring that it is not the computer that dictates the mode of interaction, but rather the needs of the users who are placed at the centre when developing technologies intended for human use. This approach is known as participatory or human-centred design (ISO, 2019). Its core principle: people affected by a system (primarily the users themselves) are actively involved in its development – from defining the problem, through gathering requirements, to evaluating how well a system actually works and is accepted.
A particular challenge in the case of autistic people was the diversity of the target group: since autism affects everyone differently, it makes little sense to develop technologies for individual people. Instead, we had to look for recurring classes of problems for which solutions could be offered to as broad an audience as possible.
A further challenge lay in how to approach the target group itself: autistic people’s willingness to take part in the project depended heavily on how sensitively we, as a research team, could handle the subject – a sensitivity we could only really develop through exchange with the target group in the first place. We resolved this circularity by first seeking professional expertise before intensifying direct contact with the target group, for example through close collaboration with the SRH Berufsbildungswerk Dresden, with specialists in curative education, and with the autism outpatient clinic at the Dresden University Hospital.
On this basis, we conducted various studies: observational studies in the training environments of the vocational training centre, interviews with trainers and trainees, and conversations with people at different stages of life. Valuable input also came from the AUT-1A and AUT*CIA projects run by the Berufsbildungswerk St. Franziskus Abensberg and the Munich University of Applied Sciences, which examined, among other things, differences between male and female autism in the workplace context (Reich & Dalferth, 2023).
A particular quirk of timing shaped the project: the planning phase fell before the emergence of ChatGPT, yet by the time the project started, a technology had suddenly become available that opened up new perspectives on assistive systems. Remarkably, it was not only we who worked with artificial intelligence (AI) in the project – many of the autistic people we interviewed were already using AI tools independently to support themselves in daily life. A clear signal of the demand for technical support.
One thing is essential to us in this context: the project was never solely about helping autistic people into employment. Work is closely linked to self-realisation, self-efficacy, the feeling of being needed, and to dignity and independence – aspects that go far beyond the economic perspective and substantially shape quality of life. In the understanding of our project, therefore, support in the work context should be seen as support for a better quality of life, not as promoting the adaptation of autistic people to a neurotypical environment.
From the listening phase described above, we derived three central questions: which challenges are actually relevant in everyday training and working life? What interactions exist between personal factors, environment and tasks? And what requirements does this create for technical, social and organisational support?
Methodologically, we drew on interviews with autistic people with several years of professional experience, interviews with neurodivergent students at TU Dresden, observations in office-related training areas of the vocational training centre, and several work placements for trainees – a particularly valuable experience that let us witness first-hand what it means to work together as a team. This was supplemented by regular exchange with our project advisory board, in which Autismus Deutschland e. V. also played an important role. The result was a catalogue of more than 90 concrete challenges, from which three central problem areas emerged, as already mentioned at the outset: communication and social interaction, sensory perception and processing, and time and task management – plus additional overarching influencing factors such as stress tolerance, need for routine, ability to concentrate, motivation and social acceptance within the environment. A key tool for the further work was also a set of five personas developed over the course of this intensive engagement: not a single stereotypical description, as is usually customary for personas, but a coherent set of different profiles that honours the diversity of autistic life realities while also serving as a shared foundation for development (Engel et al., 2025).
Communication and social interaction
An instructive finding from the requirements analysis was that directly asking about particularly burdensome everyday challenges often led to contradictory or vague answers. When asked directly, many autistic people were unable to name clear problems, but described a diffuse sense of everyday difficulty. It was only through in-depth interviews that it became apparent that these difficulties usually arise from the interplay of several areas – in particular communication and interpersonal interaction, external stimuli, and the organisation of one’s own routines and priorities.
Social interaction is a key area of concern here: dealing with colleagues, managers, or customers. Both the mode of communication – for example, one-to-one conversation, group situations such as meetings, or presenting to an audience – and the particular setting play a role: in-person conversations differ markedly from video calls, phone calls, or written communication via email and chat. Each of these constellations brings its own challenges, for instance regarding non-verbal cues, tone of voice, or potential distractions. This is compounded by the fact that the different settings offer varying degrees of scope for withdrawal when overwhelmed, and such withdrawal is not accepted in every social context. There is also frequently a lack of the necessary understanding among colleagues.
Concretely, these challenges manifest as misunderstandings caused by indirect or implicit communication, unrecognised non-verbal signals, being overwhelmed by spontaneous rather than preparable conversational situations, and pressure of expectation regarding eye contact or small talk. A widespread phenomenon is also so-called masking – the short-term adaptation to neurotypical behavioural norms, which can lead to exhaustion eventually and is therefore often a cause of withdrawal or job change (Alaghband-rad et al., 2023).
Within the project, we developed approaches to solutions for both written and face-to-face communication. Written communication focused on email, since it is particularly often used in everyday office life to assign tasks and responsibilities, and growing email threads can create additional complexity. That said, the concepts developed can in principle also be transferred to other channels such as chat. Email is often preferred over phone calls, yet frequently contains implicit expectations, irony or metaphor, and also makes it harder to prioritise and grasp the overall context. It is precisely here that AI systems, particularly those based on large language models, can potentially offer meaningful support – provided they are used in a data-protection-compliant manner within a business context.
Supporting email work with data-protection-compliant AI. The system developed in the project, the EmailAnalyzer (Gollasch et al., 2025), analyses incoming emails using three views. The first view extracts the tasks and deadlines it contains and presents them in a clear, hierarchical list. The second view identifies and explains linguistic devices such as metaphor or irony – deliberately with a direct reference back to the original passage of text, so that possible AI errors can be critically checked. The third view supports the drafting of replies: many autistic employees rely on prepared email templates for standard situations, which is practical but not very flexible. The system therefore helps find a clear thread and produce a complete, well-structured reply, without dictating a particular content direction, though it does provide structural support.
By default, the system runs on a locally executed AI model and is therefore independent of external providers, meaning no sensitive data must be sent over the internet. However, local AI models – at least at present – are still somewhat limited in output quality, so in some cases it may make sense to incorporate external AI systems approved by the company. In addition, not every office workstation has sufficient computing power for local models, though this is likely to become less of an issue in the coming years.
Augmenting support in conversational situations. For face-to-face, spoken communication, we developed a concept building on the same analytical approaches, but one that additionally needs to work in real time: support during a conversation is only helpful if it is immediate and unobtrusive. To this end, we experimented with augmented and virtual reality (AR/VR) glasses that can display additional information about the conversational situation, for example to help identify people or provide conversational context. Current AR and VR glasses are still limited and not very comfortable for practical use, but new, discreet glasses with integrated cameras, microphones and speakers, such as Meta AI Glasses, already show that environmental information can be captured and conveyed in a far more everyday-friendly way than before. Transmission via smartwatches is also conceivable here.
Sensory perception and processing
Alongside communication, it is specifically the perception of the environment that shapes the working day for many autistic people. Sensory perception and processing form a particularly complex and broad field within the challenges faced by autistic people. This is primarily because virtually every sense can be affected: visual stimuli such as brightness or fast movement, auditory stimuli such as background conversations, loud or particularly quiet, hard-to-identify noises, but also olfactory stimuli such as certain smells. Factors such as the presence of many people in the immediate vicinity can likewise trigger challenging situations. The variety of possible stimuli and their combinations is enormous (American Psychiatric Association, 2013; Ríos Llamas et al., 2026).
Overload often occurs when a particular stimulus in a specific context absorbs so much attention that it becomes almost impossible to concentrate on the task at hand. Stimuli thus become sources of distraction that not only reduce work performance but can also cause tangible stress, or even have far more serious effects. It is often not a single stimulus but the interplay of several stimuli, together with additional contextual factors, that leads to overload. Not least, a person’s day-to-day form also plays a role in how sensitive they are to particular stimuli on a given day.
This considerable variability also poses challenges for the development of technical solutions: a single solution covering the entire spectrum is hardly realistic. A sensible approach, therefore, is to build on devices that are already established and easy to adopt – such as noise-cancelling headphones, which are already widely used in working life to shield against auditory stimuli. Their drawback: auditory isolation also makes the communication with colleagues and managers that is necessary in daily work more difficult, which limits the acceptance of such headphones in the office.
Office-compatible noise suppression. This is where the system developed in the project, SmartHeadphones (Schmalfuß-Schwarz et al., 2024b, 2024a), comes in: it works similarly to voice assistants in the home, which respond to a specific activation phrase. While auditory isolation is ensured under normal conditions, the system allows social interaction to be enabled selectively when needed. Ultimately, this is about striking a balance between reducing stimuli on the one hand and the ability to carry out professional duties – including necessary communication – on the other.
A key component here is the ability to individually define when an interaction should be initiated. Unlike the mostly fixed activation phrases of commercial voice assistants, the activation phrase for SmartHeadphones is freely customisable – for example, one’s own name. When it is spoken, the headphones automatically switch into a transparency mode, or the user receives a notification and can decide themselves whether interaction is wanted at that moment.
In addition, we integrated a concept that lets users signal outwardly whether they are currently open to conversation – a so-called busy monitor, modelled on busy-light systems available either as standalone devices or as part of telephony headsets. This signal is aimed primarily at colleagues and helps ensure that social interaction is not initiated unbidden, but can be consciously invited.
Early detection of potentially stressful environmental situations. Another important aspect of sensory perception is the early detection of potentially stressful situations. Environmental variables such as temperature, brightness, air pressure or noise level can significantly influence how stimuli are perceived. This is the focus of the system developed in the project, ELVI (Environmental Level Visualization Interface): it functions as a kind of early-warning system that records relevant environmental variables and gives early warning of possible overload situations. This allows those affected to respond proactively, rather than having to deal with overload only once it has already occurred. Since individual challenges and stimulus profiles vary considerably, the central insight from this thematic area is as follows: context-sensitive and adaptive technology is needed to reduce and make controllable the burden of sensory stimuli in a targeted way, without restricting users’ self-determination. Support without paternalism is the decisive benchmark here (Jameson, 2008; Shakespeare & Watson, 2001).
Time and task management
The same principle – support that leaves room for manoeuvre rather than prescribing – also applies to the third major thematic area: the organisation of time and tasks. The third key dimension in which people on the autism spectrum experience professional challenges concerns the execution of work tasks itself: specifically, how personal resources are used to complete tasks on time and to an appropriate standard. An important factor here is how routine, plannable and complex tasks are, and how much flexibility they demand of employees. By “need for flexibility” we mean the property of tasks to change dynamically in priority and character, as opposed to highly routine, well-plannable tasks of low complexity. Which specific tasks arise in a given profession naturally depends on the field concerned and lies outside the scope of influence of technical assistive systems. What was decisive for our project, rather, was the question of how autistic people allocate their personal resources, prioritise tasks, and situate them in time.
A commonly described challenge is so-called time blindness – that is, the difficulty of realistically estimating durations and prioritising tasks accordingly against the time available (Jurek et al., 2019). This is often compounded by a strong reliance on routines and visual structures. The motivation to carry out tasks can also be moderately different for autistic people compared with neurotypical people. A further risk is so-called hyperfocus: a state of particularly intense concentration that, while enabling especially effective work, can lead to the neglect of basic needs such as breaks for recovery or even eating, and ultimately to overwork. Finally, task complexity is also a hurdle: breaking complex tasks down into manageable sub-steps can be difficult for autistic people. At the same time, this is a skill that is usually indispensable in everyday work, for example to be able to interrupt tasks at sensible points and bring the working day to an orderly close.
Individualised versions of established time-management techniques as a starting point. To develop solutions, we first drew on classic concepts from time and project management. The real difficulty, however, lies in selecting the right method in each case and adapting it individually. This can create a structure that genuinely provides a sense of security and makes it easier to cope with an increased need for flexibility. In the project, we therefore developed a system that modularly combines several established planning techniques – such as the Pomodoro Technique, timeboxing, the Eisenhower Matrix, Getting Things Done, or Kanban – and makes them individually adaptable.
A spiral view instead of tables for perceiving time. In addition, we developed a tool that enables an entirely different way of perceiving time. Whereas most of us represent appointments in tabular form – for instance in a classic calendar or in digital tools such as Outlook, with one column per day – we implemented a concept that docks onto existing calendars and instead represents time radially, similar to a clock. Besides clocks, this form of representation is also reminiscent of the Time Timer, which is used, among other things, in therapeutic contexts (Grey et al., 2009; Hallez & Vallier, 2025). Concretely, the display is arranged like an Archimedean spiral: time runs from the inside outwards in a snail-shell-like structure. This representation may seem unfamiliar at first, but it allows a different, often more intuitive way of grasping elapsed and available time, of recognising recurring patterns in daily life, and of deliberately scheduling free space such as regular breaks (Dragicevic & Huot, n.d.).
Literally tangible interaction. We additionally combined this system, called AgendaHub in the project, with a tactile device known as the AutiPuck. This is a device that can be held in the hand like a small object, or placed on the desk, to signal, for example, which phase of work one is currently in – such as focused individual work, a meeting, or a break. At the same time, the AutiPuck has a small screen that displays notifications and upcoming appointments. This creates an interplay of several devices for representing time, planning tasks, and connecting to existing digital systems such as Microsoft Outlook, which is widely used in workplace settings. One aspect that mattered a great deal to us throughout the project: the systems developed are not intended exclusively for people on the autism spectrum, but also offer added value to neurotypical employees. This means the tools are, in principle, available to the whole team, which increases acceptance within the workforce and frees the systems from any potentially stigmatising character.
Beyond the prototypes: raising awareness and providing guidance
Of course, technical systems can never be the sole form of support in daily life, but only one building block towards it. The intensive engagement with autism in the context of vocational training and professional practice over the course of the project also generated insights that go far beyond the development of specific assistive systems. For example, the personas developed proved valuable not only as a basis for technology development, but also as a valuable communication tool in a guidance context – for instance, to raise general awareness of the challenges faced by autistic people. It was precisely this undogmatic, nonmedical approach that our project advisory board particularly highlighted and praised as an important outcome.
Translating individual needs into workplace requirements. This gave rise to the idea for a standalone guidance tool, the Workplace Compass: a dynamic, interactive questionnaire that captures individual challenges and wishes in the work context and derives concrete recommendations for suitable workplace design from them. The tool can be used in several ways: an employee on the autism spectrum can complete the questionnaire and pass on the resulting recommendations to their employer, who can then adapt the working environment accordingly. At the same time, the tool is also suitable for careers guidance and job placement, to help identify suitable employment opportunities more effectively. This is because it is often difficult for people to clearly articulate their own needs regarding an ideal workplace. Here, the Workplace Compass creates the necessary transparency and facilitates communication.
With this tool, we deliberately went beyond pure assistive technologies and also considered adjacent influencing factors. Alongside the actual working environment, the path leading up to it – that is, vocational training and study – also came into focus. The advantage of all the systems developed in the project, and of comparable technologies, is that they can already be introduced during the qualification phase and later carried over into the actual workplace. This creates a level of support that has hardly been addressed so far: many existing offerings rely predominantly on coaching or training in how to cope with challenging situations, but do not relieve those affected of continuing to be directly exposed to these challenges in daily life. The technologies developed in the project create additional value here by offering, alongside personal coaching, immediate, practical support that can provide genuine relief.
What mattered throughout was never to create paternalism through technology, but rather to preserve the greatest possible freedom and flexibility. The systems developed are not tied to mandatory use, but are available to those affected entirely according to individual need.
Conclusion and outlook
A particular advantage of the technologies developed is that they were created without any commercial self-interest and have been published as fully and freely accessible open source. The systems are naturally at different levels of maturity and are so far available as prototypes, which is typical for research projects. This now allows third parties to pick up the results with a low barrier to entry, develop them further, and integrate them into existing systems. The Workplace Compass guidance tool and the personas developed are already fully available today.
Overall, the research project was highly successful: over a good three years, 28 studies were conducted with a total of 167 participants – both autistic people themselves and professional experts from the surrounding field (Technische Universität Dresden, 2026). The impact was also felt within the scientific community: we offered workshops on neurodivergence and autism at three international HCI conferences. A notable development emerged over the course of the project: while at the first workshop in 2023 many participants’ focus was still more strongly on children and on ADHD, by the final workshop at the end of 2025 autistic researchers themselves were actively contributing and placing autism in adulthood at the centre – a clear sign of growing awareness within just three years. The closing event at the SRH Berufsbildungswerk Dresden likewise confirmed the high relevance of the subject: more than 120 people from vocational rehabilitation and training took part and were able to experience the prototypes developed live, with palpable enthusiasm.
The project has also contributed to raising awareness on many levels: through the work placements created, through close collaboration with self-advocacy organisations such as Autismus Deutschland e. V., whose involvement on the project advisory board provided valuable input, and through the active participation of autistic people themselves. Topics such as neurodiversity and AI-supported assistance remain relevant – exchange within the EUCAP-AIRA project, which specifically addresses AI and autism, likewise provided important impetus for future potential and an overview of ongoing activities beyond Germany’s borders (AIRA Artificial Intelligence and the Rights of Autistic people, 2025).
We would like to emphasise clearly how important close exchange with organisations such as Autismus Deutschland e. V. is for the further development of such technologies. A key success factor of human-centred design lies in taking the term literally: bringing the right people together in the right place, putting them at the centre, connecting them, and working together on helpful solutions.
Literature
AIRA Artificial Intelligence and the Rights of Autistic people. (2025, Februar). Artificial Intelligence and Autism: Toolkit for Activists. EUCAP European Council of Autistic People.
Alaghband-rad, J., Hajikarim-Hamedani, A., & Motamed, M. (2023). Camouflage and masking behavior in adult autism. Frontiers in Psychiatry, 14, 1108110. https://doi.org/10.3389/fpsyt.2023.1108110
American Psychiatric Association. (2013). Diagnostic and Statistical Manual of Mental Disorders (DSM-5®). American Psychiatric Publishing. http://ebookcentral.proquest.com/lib/slub/detail.action?docID=1811753
Bader, M., Labruier, M., Proft, J., & Vogeley, K. (2018). Menschen mit Autismus im Arbeitsleben Informationen und Handlungsempfehlungen.
Dragicevic, P., & Huot, S. (o. J.). SpiraClock: A continuous and non-intrusive display for upcoming events.
Engel, C., Gollasch, D., Branig, M., Schmalfuß-Schwarz, J., & Weber, G. (2025). Personas of Autistic Office Workers: A Foundation for Developing Effective Assistive Technologies. In M. Antona & C. Stephanidis (Hrsg.), Universal Access in Human-Computer Interaction (Bd. 15780, S. 36–55). Springer Nature Switzerland. https://doi.org/10.1007/978-3-031-93848-1_4
Gollasch, D., Branig, M., Dähne, H., & Weber, G. (2025). Automated Task Extraction from Emails: Aiding Autistic Individuals in the Workplace. In M. Antona & C. Stephanidis (Hrsg.), Universal Access in Human-Computer Interaction (Bd. 15780, S. 343–359). Springer Nature Switzerland. https://doi.org/10.1007/978-3-031-93848-1_23
Grey, I., Healy, O., Leader, G., & Hayes, D. (2009). Using a Time TimerTM to increase appropriate waiting behavior in a child with developmental disabilities. Research in Developmental Disabilities, 30(2), 359–366. https://doi.org/10.1016/j.ridd.2008.07.001
Hallez, Q., & Vallier, V. (2025). Time on Their Side: How Visual Timers Affect Anticipatory Anxiety, Performance, and On-Task Behavior in Elementary Math Assessments. European Journal of Investigation in Health, Psychology and Education, 15(12), 243. https://doi.org/10.3390/ejihpe15120243
ISO, I. O. for S. (2019). ISO 9241-210:2019 – Ergonomics of human-system interaction Part 210: Human-centred design for interactive systems.
Jameson, A. (2008). Adaptive Interfaces and Agents. In The Human-Computer Interaction Handbook – Fundamentals, Evolving Technologies and Emerging Applications (2. Aufl.).
Jurek, L., Longuet, Y., Baltazar, M., Amestoy, A., Schmitt, V., Desmurget, M., & Geoffray, M.-M. (2019). How did I get so late so soon? A review of time processing and management in autism. Behavioural Brain Research, 374, 112121. https://doi.org/10.1016/j.bbr.2019.112121
Reich, K., & Dalferth, M. (2023). Vom Vermittlungsparadigma hin zur nachhaltigen Unterstützung von Unternehmen bei der Eingliederung von Menschen mit hochfunktionalen Autismus – Ergebnisse einer Arbeitgeberbefragung. Die Rehabilitation. https://doi.org/10.1055/a-1998-6630
Ríos Llamas, C., Camacho Vega, D. O., & Delgadillo Ramos, M. G. (2026). Light and sound hypersensitivity in autism spectrum disorder: A systematic review focusing on age and gender bias. Frontiers in Psychiatry, 17, 1771956. https://doi.org/10.3389/fpsyt.2026.1771956
Schmalfuß-Schwarz, J., Gollasch, D., Engel, C., Branig, M., & Weber, G. (2024a). Customized Wake-Word Integration in ANC-Enabled Headphones: Enhancing Assistive Technology for Noise-Sensitive Employees in Office. Proceedings of Mensch Und Computer 2024, 533–537. https://doi.org/10.1145/3670653.3677489
Schmalfuß-Schwarz, J., Gollasch, D., Engel, C., Branig, M., & Weber, G. (2024b). Open Sesame! Use of Headphones at Work Considering Social Acceptance. In K. Miesenberger, P. Peňáz, & M. Kobayashi (Hrsg.), Computers Helping People with Special Needs (Bd. 14751, S. 420–429). Springer Nature Switzerland. https://doi.org/10.1007/978-3-031-62849-8_52
Shakespeare, T., & Watson, N. (2001). The social model of disability: An outdated ideology? In Research in Social Science and Disability (Bd. 2, S. 9–28). Emerald (MCB UP ). https://doi.org/10.1016/S1479-3547(01)80018-X
Technische Universität Dresden. (2026). Projekt-Abschlussbericht, Sachbericht zum Projekt Unterstützung des Arbeitsalltags von Menschen mit Autismus mittels kontextsensitiver Anwendungen zur Kommunikation und zum Aufgabenmanagement (AutARK) [Projekt-Abschlussbericht]. Technische Universität Dresden. https://www.rehadat-forschung.de/export/sites/forschung-2021/lokale-downloads/BMAS/FO126085_Abschlussbericht.pdf
