Skip to content
Technical University of Munich

BMBF Research Hub: 6G-Life

6G-life is a BMBF-funded research hub, which will drive cutting-edge research for 6G communication networks with a focus on human-machine collaboration. 6G-life provides new approaches for sustainability, security, resilience and latency and will sustainably strengthen the economy and thus digital sovereignty in Germany.

Motivation

TU Dresden and the Technical University of Munich have joined forces to form the 6G-life research hub to drive cutting-edge research for future 6G communication networks with a focus on human-machine collaboration. The merger of the two universities of excellence combines their world-leading preliminary work in the field of Tactile Internet in the Cluster of Excellence CeTI, 5G communication networks, quantum communication, Post-Shannon theory, artificial intelligence methods, and adaptive and flexible hardware and software platforms.

6G-life will significantly stimulate industry and the startup landscape in Germany through positive showcase projects and thus sustainably strengthen digital sovereignty in Germany. Test fields for two use cases will drive research and economic stimulation. The goal is to create at least 10 new startups through 6G-life in the first four years and to involve at least 30 startups. 6G-life will significantly contribute to the creation of a skilled workforce. In addition, 6G-life has set itself the task of accompanying the population in the digital transformation and thus making a contribution to society.

Research focus at ITR

Our contribution to 6G-Life is the development of theoretical foundations for networked control of safety-critical systems, with rehabilitation robotics as a representative demonstrator. The work combines control theory, machine learning, and distributed computation with the aim of ensuring stability, safety, and adaptability under communication constraints.

Communication-aware Control Architectures

We design control frameworks where immediate safety is ensured locally on the robot, while computationally demanding tasks such as optimisation and estimation are placed at the network edge. Longer-term adaptation and coordination are carried out in the cloud. Examples include in-network estimation of human motion from wearable sensors and shared control strategies for hybrid systems that combine exoskeletons with functional electrical stimulation.

QoS-informed Adaptation

Our work integrates information on latency, jitter, and packet loss directly into control design. This enables controllers to adapt sampling rates and gains, or to switch between networked and local backup strategies. Approaches based on control barrier functions and delay-aware learning ensure that stability and safety are maintained even when network quality changes.

Data Prioritisation for Learning and Control

We develop scheduling strategies that distinguish between time-critical control signals and less urgent data for monitoring or personalisation. Methods such as value-based scheduling and selective information sharing reduce network load while guaranteeing timely delivery of safety-critical data. In rehabilitation systems, this ensures that stimulation and sensing information is preserved while background data can be transmitted flexibly.

Rehabilitation as Demonstrator

The theories are validated in robotic rehabilitation systems where strict real-time guarantees and personalised human models are essential. Hybrid exoskeleton–FES devices and soft wearable gloves provide representative test cases for showing how 6G-enabled in-network computing can support safe, mobile, and user-specific assistance.

Relevant topics:

The placement of control components across robot, edge, and cloud raises fundamental questions for networked control theory. The figure illustrates different possible allocations of modules such as virtual impedance elements, admittance controllers, and reference generators. Each configuration highlights a distinct research challenge.

  • Partitioning of control loops
    A central problem is to determine which elements must remain local for guaranteed stability and which can be safely migrated to the network. This involves analysing the passivity properties of subsystems, deriving conditions for compositional stability, and quantifying how delay and packet loss affect interconnected modules.
  • Trade-off between robustness and flexibility
    Architectures with more local control achieve robustness to network variations, but limit the complexity of algorithms that can be used. Offloading to the network increases computational flexibility but introduces delay-sensitive coupling. We investigate design principles for balancing these opposing requirements.
  • Compositional design of distributed controllers
    When controllers are split across multiple nodes, the interconnection must be stable despite asynchronous updates. This motivates research on distributed passivity theory, network-induced delay modelling, and layered stability guarantees that ensure safe behaviour even when parts of the controller are remote.
  • Role of 6G in-network computing
    Emerging features such as guaranteed low-latency links, prioritised data channels, and computational resources embedded in the network provide new opportunities. Our research addresses how these capabilities can be formally integrated into control design, enabling architectures that were previously impractical under wireless communication constraints.

Overall, the research aims to establish general principles for allocating observers, estimators, and controllers across distributed computational layers. These principles are validated in robotic demonstrators but are intended to be broadly applicable to networked control systems in safety-critical domains.

The performance of networked control systems is directly influenced by the quality of the underlying communication. Unlike classical control setups, latency, jitter, and packet loss can vary dynamically in wireless networks. QoS-informed adaptation makes these network properties explicit inputs to the control design, leading to new research directions.

  • Real-time QoS feedback
    • Controllers can adapt sampling rates, controller gains, or prediction horizons based on live QoS metrics.
    • High-quality links allow more aggressive control with shorter update intervals, whereas degraded conditions require conservative settings to maintain stability.
  • Safe switching mechanisms
    • Systems may need to switch between networked controllers and local fallback strategies.
    • Theoretical work focuses on deriving conditions that guarantee stability and passivity during such mode transitions, ensuring that safety is not compromised.
  • Prescribed-time safety guarantees
    • When dynamics are uncertain, control barrier functions combined with data-driven models can enforce safety constraints within a finite time window.
    • This extends beyond asymptotic guarantees and is particularly relevant for time-critical tasks where recovery must be achieved quickly.
  • Delay-aware learning integration
    • Online learning methods, such as Gaussian process regression, are effective for adaptation but introduce computational delays.
    • We investigate how to explicitly model these delays, derive accuracy–delay trade-offs, and design event-triggered learning schemes that update only when stability margins allow.
  • Coupling with communication technology
    • 6G networks are expected to provide QoS monitoring services at the device level.
    • Research questions include how to formalise the interface between QoS indicators and control algorithms, and how to guarantee that adaptation is both fast and verifiable.

The goal of this line of research is to establish a framework in which network quality is not treated as a disturbance but as structured information that can be exploited to preserve safety and enhance performance in distributed control systems.

Example Application: Kalman Filter: Prediction + Correction

  • Robust against connection loss and packet dropout
  • Output limitation for stable estimation
  • Applicable to networked estimation tasks such as human pose tracking, motion prediction, and other real-time sensing scenarios

In networked robotic systems, the communication channel must carry both time-critical control signals and less urgent data for monitoring, diagnostics, or learning. Treating all data streams equally leads to congestion and undermines real-time guarantees. Data prioritisation focuses on separating these streams and assigning resources according to their control relevance.

  • Value of Information in control
    • Not every data packet contributes equally to closed-loop performance.
    • We study real-time metrics that quantify the marginal utility of data for control stability and accuracy. These metrics are used to decide which information is transmitted immediately and which can be buffered or aggregated.
  • Scheduling for mixed traffic
    • Control-critical signals must meet strict latency and jitter bounds, while learning-related data can tolerate relaxed deadlines.
    • Our research investigates scheduling algorithms that guarantee hard deadlines for critical data while opportunistically using spare capacity for non-urgent traffic.
  • Distributed learning with selective communication
    • In multi-agent systems, communication overhead can be reduced by selective data exchange.
    • Agents can decide when and with whom to share predictions or measurements, based on estimated reliability and the accuracy of their own models. This reduces bandwidth requirements without sacrificing prediction quality.
  • Interaction with network services
    • 6G features such as network slicing, traffic prioritisation, and in-network buffering open new possibilities.
    • We analyse how these services can be co-designed with control requirements, so that network-level resource allocation aligns with control-theoretic guarantees.
  • Balancing safety and long-term adaptation
    • Prioritisation must preserve stability while still enabling data-driven model improvement.
    • Research questions include how to ensure that delayed or dropped learning data does not bias adaptation, and how to couple online learning with strict timing requirements of control loops.

The outcome of this research is a principled framework where data is treated as a heterogeneous resource: critical for safety in some cases, informative for adaptation in others. By explicitly modelling the control relevance of different streams, it becomes possible to maintain real-time guarantees while supporting efficient learning and monitoring in distributed robotic systems.

We have advanced networked control methods through demonstrators that operate with patient cohorts in clinical contexts. These include wearable inertial measurement units for continuous symptom monitoring in Parkinson’s disease, and hybrid upper-limb assistance systems combining exoskeleton support with functional electrical stimulation and multimodal sensing such as EMG. Both represent networked systems in which sensing, estimation, and control are distributed across devices and external computation.

These demonstrators highlight two complementary aspects of clinical translation. On the one hand, continuous monitoring applications illustrate how networked sensing and learning can provide long-term information on disease progression. On the other hand, rehabilitation devices show how distributed controllers can safely support patient movements in real time. Together, they provide concrete use cases for testing how 6G features such as in-network computing, guaranteed latency, and data prioritisation can enable clinically relevant performance outside laboratory conditions.

Videos

ITR team members

Selected publications

2025

  • Donié, Cedric; Das, Neha; Endo, Satoshi; Hirche, Sandra: Estimating Motor Symptom Presence and Severity in Parkinson's Disease from Wrist Accelerometer Time Series Using ROCKET and InceptionTime. Nature Scientific Reports, 2025 more… BibTeX Full text ( DOI ) Full text (mediaTUM)

2024

  • Kavianirad, Hossein; Missiroli, Francesco; Endo, Satoshi; Masia, Lorenzo; Hirche, Sandra: Toward Dexterous Hand Functional Movement: Wearable Hybrid Soft Exoglove-FES Study. 2024 10th IEEE RAS/EMBS International Conference for Biomedical Robotics and Biomechatronics (BioRob), 2024 more… BibTeX Full text ( DOI ) Full text (mediaTUM)
  • Tzu-Yuan Huang, Sihua Zhang, Xiaobing Dai, Capone, Alexandre, Todorovski Velimir, Sosnowski Stefan, Sandra Hirche: Learning-based Prescribed-Time Safety for Control of Unknown Systems with Control Barrier Functions. IEEE Control System Letter, 2024 more… BibTeX Full text ( DOI ) Full text (mediaTUM)
  • Yang, Zewen and Dai, Xiaobing and Dubey, Akshat and Hirche, Sandra, Hattab, Georges: Whom to Trust? Elective Learning for Distributed Gaussian Process Regression. 2024 more… BibTeX Full text (mediaTUM)

2023

  • Capone, Alexandre; Jiao, Junjie; Zarei, Mostafa; Zhang, Shiqi; Hirche, Sandra: Robust H∞ Consensus for Homogeneous Multi-agent Systems with Parametric Uncertainties. 2023 American Control Conference (ACC), 2023 more… BibTeX Full text ( DOI ) Full text (mediaTUM)
  • Di Liu, Simone Baldi,Sandra Hirche: Collision Avoidance in Longitudinal Platooning: Graceful Degradation and Adaptive Designs. IEEE Control Systems Letters , 2023 more… BibTeX Full text (mediaTUM)
  • Gao, Yuan; Jiao, Junjie; Hirche, Sandra: H2 suboptimal leader-follower consensus control of multi-agent systems. 22nd IFAC World Congress (IFAC WC 2023), 2023 more… BibTeX Full text (mediaTUM)
  • Gao, Yuan; Jiao, Junjie; Hirche, Sandra: H2 suboptimal containment control of multi-agent systems. 2023 American Control Conference (ACC), 2023 more… BibTeX Full text ( DOI ) Full text (mediaTUM)
  • Kavianirad, Hossein; Forouhar, Moein; Sadeghian, Hamid; Endo, Satoshi; Haddadin, Sami; Hirche, Sandra: Model-Based Shared Control of a Hybrid FES-Exoskeleton: an Application in Participant-Specific Robotic Rehabilitation. 2023 International Conference on Rehabilitation Robotics, ICORR 2023, 2023 more… BibTeX Full text ( DOI ) Full text (mediaTUM)
  • Xiaobing Dai, Armin Lederer, Zewen Yang, Sandra Hirche: Can Learning Deteriorate Control? Analyzing Computational Delays in Gaussian Process-Based Event-Triggered Online Learning. 2023 more… BibTeX Full text (mediaTUM)
  • Xiaobing Dai, Huanzhuo Wu, Siyi Wang, Junjie Jiao, Giang T Nguyen, Frank HP Fitzek, Sandra Hirche: Fast IMU-based Dual Estimation of Human Motion and Kinematic Parameters via Progressive In-Network Computing. 2023 more… BibTeX Full text (mediaTUM)

Project information

  • Participating Universities
    • Technische Universität Dresden
    • Technical University of Munich
  • Start date: 15th August, 2021
  • Funding: 70 Million € for 4 years
  • Number of principal investigators: 60
  • Coordinators:
    • Prof. Dr.-Ing. Dr. h.c. Frank Fitzek (TU Dresden)
    • Prof. Dr.-Ing. Dr. rer. nat. Holger Boche (TUM)
  • Project website: https://6g-life.de/

Funding program

The BMBF’s funding measure “6G Research Hubs; Platform for Future Communication Technologies and 6G” is part of the implementation of the Federal Government’s Future and Economic Stimulus Package and is funded with around 200 million euros from the Future Package. With its research on 6G, Germany is setting itself the goal of assuming a leading role among the world’s top technology providers and helping to shape technological change at an early stage. This also contributes to the implementation of the German government’s High-Tech Strategy 2025.

The authors acknowledge the financial support by the Federal Ministry of Education and Research of Germany in the programme of “Souverän. Digital. Vernetzt.”. Joint project 6G-life, project identification number: 16KISK001K.

To top