Research Program

RA1 develops methodologies for designing interactive systems that are safe, explainable, and user-centered. It investigates human-in-the-loop co-design approaches that integrate user characteristics, preferences, and feedback into engineering processes.
The research combines formal methods, behavioral modeling, and experimental interaction studies to support adaptive and reliable system design.
Applications focus on medical and assistive robotics, where close human–system collaboration is essential.
RA1 contributes to coHu by providing methodological foundations for human-centered design, bidirectional communication, and mutual understanding.
The research area is structured into three subprojects addressing safe interfaces, participatory design methods, and personalized cognitive robotics.

RA1-SP1 – Co-design of safe and explainable human-system interfaces

This subproject develops methods for designing interactive interfaces that are understandable to users while ensuring formal safety. It addresses the challenge of increasing system adaptability in safety-critical domains such as healthcare and robotics.
The approach combines human-in-the-loop experiments, behavioral modeling, and formal verification to assess safety and interaction properties during design.
A key objective is explainability, enabling transparent communication of system intentions, states, and actions.

Doctoral Project – Co-verification for Provably Safe Human-System Interaction

• Develops computational models of human behavior for formal verification of interactive systems
• Bridges the gap between precise system models and variable human behavior
• Learns behavioral models from user data and integrates them into verification pipelines
• Focuses on intent prediction and early detection of unsafe situations
• Validated in assistive robotics and wearable technologies

 

RA1-SP2 – Human-in-the-loop methods

This subproject develops methods to systematically integrate users into interactive system design. It addresses the limited integration of user needs and experiences into engineering processes.
It enables continuous user involvement through co-design methods combining engineering, human sciences, and UX research.
Immersive VR/AR environments are used to evaluate interaction concepts early in development.

Doctoral Project – Integration of Co-design and Human-in-the-loop Design Methods

• Builds a framework combining engineering constraints and user-centered criteria (comfort, trust, usability)
• Integrates user data from studies, questionnaires, and physiological signals into design decisions
• Evaluates methods in wearable robotics, especially ankle exoskeletons
• Enables continuous user feedback throughout system development
• Shifts user input from final validation to core design stage

RA1-SP3 – Personalized Cognitive Robotics

This subproject develops adaptive robotic systems that act as intelligent partners capable of understanding user intentions and adapting to individual needs.
It focuses on learning-based, multimodal and shared-autonomy approaches to improve human–robot collaboration.
Applications target robot-assisted surgery, where prediction and real-time adaptation are critical.

Doctoral Project – Human-Robot Collaboration through Anticipation of Surgical Intent

• Learns surgical strategies from demonstrations and observed procedures 
• Uses video and machine learning to infer surgeon intent from motion and context 
• Integrates intent prediction into robotic control for proactive assistance 
• Enables real-time adaptive collaboration in surgical environments 
• Validated in laparoscopic surgery scenariosLearns surgical strategies from demonstrations and observed procedures 
• Uses video and machine learning to infer surgeon intent from motion and context 
• Integrates intent prediction into robotic control for proactive assistance 
• Enables real-time adaptive collaboration in surgical environments 
• Validated in laparoscopic surgery scenarios

RA2 investigates how humans and interactive systems continuously adapt to one another during long-term interaction. It studies the biomechanical, motor, perceptual, and cognitive processes that shape co-adaptation in human–system interaction.
The research combines computational modeling, experimental studies, and adaptive technologies to understand skill acquisition, intent communication, and embodiment.
Applications focus on assistive and rehabilitative robotics, wearable systems, and surgical technologies.
RA2 contributes to coHu by explaining how mutual adaptation supports continuous interaction, effective information exchange, and shared understanding.
The research area is structured into three subprojects addressing biomechanical modeling, co-adaptation assessment, and perceptual–cognitive integration.

RA2-SP1 – Biomechanical Models for Human Interaction

This subproject develops computational models of human motion in physically collaborative tasks involving direct interaction with robotic systems. It aims to capture how humans move, coordinate, and express intent during actions such as grasping, handover, or tool use.
It combines biomechanical principles with data-driven and optimization-based modeling to infer the objectives underlying human movement.
These models enable prediction and simulation of human behavior in collaborative environments for assistive robotics design.

Doctoral Project – Human Modeling for Estimation of Motion and Intent in Collaborative Grasping Tasks

• Develops biomechanical models of hand and arm motion during human–robot collaboration
• Uses motion capture data and inverse optimal control to infer movement objectives
• Explores different levels of model complexity and contact representations
• Aims to predict human intent during collaborative grasping tasks
• Supports anticipatory robot behavior in shared manipulation scenarios

 

RA2-SP2 – Assessment and Measurement of Co-adaptation

This subproject studies how humans and assistive systems co-evolve over time during prolonged interaction. It addresses the limitations of static evaluation methods that fail to capture dynamic adaptation in real use.
SP2 develops new assessment and training protocols that explicitly integrate co-adaptation between user and system.
It combines experimental studies and adaptive paradigms to monitor changes in motor skills, behavior, and system control strategies.

Doctoral Project – Simultaneous Assessment and Training in Assistive Robotics

• Develops protocols for simultaneous assessment and training of motor skills in assistive robotics
• Studies co-adaptation in both laboratory and real-world conditions using experimental designs
• Monitors user performance and system adaptation over time
• Integrates usability and acceptance as key evaluation criteria
• Targets prosthetics, exoskeletons, and assisted surgery applications

RA2-SP3 – Human Perception and Cognition during Interaction

This subproject investigates how users perceive and cognitively integrate assistive systems during interaction. It focuses on how robotic devices can become embodied, being perceived as extensions of the body and action capabilities.
SP3 studies multisensory integration and neurocognitive mechanisms underlying learning and control of assistive devices.
It combines behavioral experiments, computational modeling, and neurophysiological measurements to analyze perception and embodiment.

Doctoral Project – Developing Brain Imaging Methods to Decode Novel Motor Programs

• Studies how new motor representations emerge during control of assistive devices
• Uses EEG and neuroimaging to track evolution of sensorimotor action schemes
• Applies machine learning to decode neural activity linked to movement intentions
• Investigates brain adaptation to prostheses and robotic limbs
• Aims to improve intuitiveness of assistive human–system interfaces

RA3 develops sensing, communication, and control technologies for real-time and reliable human–robot interaction. It investigates how human motion, intent, and environmental information can be measured, exchanged, and integrated into robotic decision-making.
The research combines advanced sensing, bidirectional communication, and predictive control to support coordinated and adaptive interaction.
Applications focus on assistive robotics, clinical environments, and safety-critical collaborative scenarios.
RA3 contributes to coHu by providing the technological foundations for interaction design, communication modalities, and mutual understanding.
The research area is structured into four subprojects addressing integrated sensing and communication, environmental perception, intent modeling, and predictive control.

RA3-SP1 – Integrated Communication and Sensing

This subproject develops tightly integrated sensing and communication systems for real-time human–robot interaction. It addresses the need for reliable information exchange and perception in dynamic environments such as healthcare and assistive robotics.
SP1 combines wireless communication and sensing technologies (e.g., radar-based perception) to simultaneously measure, process, and transmit interaction data.
The goal is to enable coordinated, safe, and efficient interaction in multi-agent settings under real-time constraints.

Doctoral Project – Fundamental Limits of Integrated Sensing and Communication

• Studies trade-offs between sensing quality and communication performance in human–robot systems
• Optimizes sensor placement, signal design, and system architecture for joint sensing/communication
• Analyzes real-time interaction constraints in assistive and medical scenarios
• Considers human perceptual and cognitive limits in system design
• Aims to define efficient and human-compatible ICAS frameworks

RA3-SP2 – Sensing for Environmental Perception

This subproject develops novel sensing methods for accurate environmental perception in human–robot interaction. It focuses on improving estimation of position and motion in dynamic, safety-critical environments.
SP2 introduces the concept of human-borne radar, where a sensor mounted on the human body exploits natural motion to enhance perception.
The approach aims to improve robustness and accuracy of shared environment understanding between humans and robots.

Doctoral Project – Human-borne radar systems for environmental perception

• Designs a body-mounted radar system using human motion for virtual aperture synthesis
• Develops trajectory estimation and reconstruction methods for non-uniform sampling
• Uses compressive sensing and non-uniform Fourier techniques for signal reconstruction
• Evaluates performance against conventional radar systems
• Targets clinical and collaborative human–robot environments

RA3-SP3 – Elucidating human intent through the optimization in the central nervous system

This subproject models human movement as the result of optimization processes in the central nervous system. It aims to identify the objectives underlying motor behavior such as effort minimization or energy efficiency.
SP3 studies how these objectives vary across individuals and tasks, and how they shape movement strategies.
The results are used to improve predictive models of human motion and enhance human–robot interaction.

Doctoral Project – Defining movement objectives in human motor control

• Identifies cost functions underlying human movement (e.g., effort, metabolic cost, kinematics)
• Uses controlled perturbation experiments to infer motor objectives
• Studies both periodic and goal-directed movements
• Validates inferred objectives in predictive motor models
• Applies findings to gait analysis and surgical robotics

 

RA3-SP4 – System Control

This subproject develops control strategies for human–robot collaboration that incorporate predicted human intent. It focuses on physical interaction tasks such as object manipulation and handover scenarios.
SP4 integrates human behavior models into Model Predictive Control (MPC) frameworks to enable anticipatory and adaptive robot behavior.
The goal is to improve safety, robustness, and naturalness of interaction in dynamic shared tasks.

Doctoral Project – Continuous intent prediction for human-robot interaction control

• Develops continuous prediction of human intent in physical interaction tasks
• Integrates force, torque, vision, radar, and gaze signals for intent estimation
• Embeds predictions into model predictive control frameworks
• Evaluates performance in cooperative manipulation scenarios
• Compares to baseline control without intent prediction