Call for Speakers 2025

Wearable neurorobots are systems designed to interpret human neural patterns and translate them into commands for wearable robotic devices, such as exoskeletons. However, the widespread adoption of neurorobotics for everyday is still limited. This project aims at redefining the neurorobotic system as a multifaceted symbiotic entity consisting of three intelligent agents---user, decoder, and robotic device. By integrating neuroscientific and robotic methodologies, the goal is to create and actively encourage shared interactions among these intelligent agents: on the one hand, the user interacts with the exoskeleton through a hybrid neural interface; on the other hand, the exoskeleton is treated as a semi-autonomous agent capable of contextualizing the user’s commands and adapt the gait trajectory to the environment.

We delve into the impact of a shoulder exoskeleton on cognitive, muscular, and oxygen consumption (VO2) expenditures. Researchers assessed the device's influence on cognitive load through the use of NASA TLX. Muscular expenditures were examined, focusing on the engagement of shoulder muscles during exoskeleton use using an electromyography (EMG). Additionally, with a gas analyzer the study measured VO2 expenditures to understand the metabolic demands of wearing the exoskeleton. Findings shed light on the potential benefits and challenges associated with shoulder exoskeletons, offering insights into their ergonomic implications and physiological effects. This multidimensional analysis contributes valuable information for the development and optimization of wearable assistive technologies, enhancing our understanding of the interplay between cognitive, muscular, and metabolic factors in exoskeleton-assisted activities.

Industrial exoskeletons are more accepted in the workplace to reduce work-related musculoskeletal disorders (WMSDs). Workers perform different activities and positions, and these actions require some versatility from the exoskeleton. Active exoskeletons provide a more effective solution because the system can be modified to modulate proper force assistance. The user now has access to different control strategies; however, what is concerning is how much of these exoskeleton domains should be open to the user? User interaction becomes challenging, and with the new advantages and power of AI tools, the exoskeleton user-interaction challenge is centered on safety and functional experience. This academic talk presents the design and assessment of an AI-driven voice user interface (VUI) to initialize and operate occupational exoskeletons.

At the University of Southern Denmark (SDU), we developed a human-inspired compliance controller, in which stiffness and damping parameters are online adapted to variable exoskeleton applications. One is assist-as-needed (AAN) control for rehabilitation by a portable (425 g) elbow exo. Another is resist-as-needed (RAN) control for decoding finger haptic patters in physical manipulation. These implementations rely only on internal actuator sensing (e.g., position), rather than expensive force/torque feedback. Our controller provides a novel way to adaptive control of physical human-exo interaction with few sensing.

The Kickstart® Walk Assist system is a neurorehabilitation device that consists of a hip belt, a multi-joint bracing system and a spring-powered Exotendon. Kickstart provides stability and assistance to lift, swing and guide the leg properly for each step without using electronics. Consecutive clinical studies demonstrated that the Kickstart exoskeletal device had a positive effect in improving walking ability in stroke survivors. In addition, the exoskeletal device could be easily integrated into current therapeutic programs to help patients enhance walking recovery both for hospital- and community-based rehabilitation training.

At the Geriatronics Research Center in Garmisch-P we are developing an Exoskeleton for future applications in rehabilitation and assistive technologies as well as to remotely control our service robot GARMI. Our exoskeleton is formed by 4 DOF and utilizes Bowden cables to transmit forces from external actuators to the wearer's limbs, enabling natural joint movement and assistance in performing tasks. However, further optimizations are required to address challenges such as cable friction and compliance to enhance the exoskeleton's overall effectiveness and usability

Providing optimal assistance is a main goal for active industrial exoskeletons, which requires estimating the payload the operator is lifting or carrying. The main problem associated with such payload estimation is that the payload is usually not directly attached to the exoskeleton: it is lifted by the operator, which in turn is supported by the exoskeleton. Due to such human intermediation, even by measuring the exoskeleton interaction forces, it is impossible to distinguish forces due to the payload and forces due to the human. By relying on EMG-driven neuromusculoskeletal models, our “delta torque” approach is able to make this distinction, leading to accurate payload estimation with low training efforts, simplified regressions and low computational time.

Through a multisite collaboration our team have been exploring the capability of the ExoMotus X2 lower limb exoskeleton to be modified for more intuitive user-controlled function, and tested in pilot trials in both healthy and stroke affected participants. Customisation is a cornerstone of rehabilitation amongst clinical populations with heterogenous neurological deficits and is enabled by open-source platforms. The Fourier Exoskeleton & Robotics Open Platform System (EXOPS) provides a means for researchers to develop and test new functionality without the overhead of building their own hardware. This presentation outlines our team’s path from device receipt, to modification, to new functionality development, and projects future changes the team plan. Our experience may serve as a template for other teams with analogous goals.

The assistance from powered lower-limb exoskeletons can extend the duration of the rehabilitation sessions for spinal cord injury and stroke patients. The help of an exoskeleton can reduce or delay the development of muscular fatigue as a risk factor for spasticity. The need to avoid the development of fatigue is crucial in rehabilitation because it negatively impacts the patient's progress. In this presentation, we will provide an overview of our approach to assessing muscular fatigue via electromyographic signal spectral analysis. This non-invasive approach aims to detect fatigue in its early stage of development and represents the first step toward a real-time fatigue detection approach that can be applied in clinical settings for the rehabilitation of neurologically impaired patients.

This talk presents a number of studies on the use of low-cost exoskeletons in physically demanding work situations of workers in a non-industrialized country. Several versions have been developed and analyzed with the aid of inertial-based monitoring technologies, EMG. Devices have been developed in the construction industry, kinematics of seated posture in ultra-heavy vehicle drivers and prolonged bipedal posture maintenance in the manufacturing industry. here is an example of one of these studies. We Design, validate and verify an ergonomic and adjustable devices for workers, due to the increase of musculoskeletal pathologies. Our devices will provide support and comfort without interfering with the task, with the objective of reducing the risk of musculoskeletal injuries, improving the health, safety of workers.

This talk presents three innovative actuation and control concepts for assistive exoskeletons. These concepts aim to reduce costs and improve exoskeleton controllability and energy efficiency at the same time. The concepts are related to (1) How to avoid oversizing actuators to obtain more lightweight, affordable, and transparent systems (2) How to enhance force controllability by mechanics in hybrid exoskeletons (3) How to improve device transparency by using non-collocated force measurement combined with low-cost accelerometers. All these three concepts are motivated by theoretical analyses, validated in real-world settings, and demonstrated on a simple upper limb exoskeleton prototype.

Delve into the blending of the unique advantages of both exoskeletons and orthoses in our presentation, “Bridging Exoskeleton & Orthosis for Improved Health Outcomes." We explore how combining the simplicity and portability of orthoses with the sophistication and usability of exoskeletons can revolutionize rehabilitation. Our focus is on the profound impact this integration has on users' ability to maintain a consistent training regimen, regardless of their location, being at rehabilitation centers or at home. Join us as we illuminate the path towards enhanced health outcomes through this innovative approach. Together, let's uncover the transformative potential of bridging traditional and advanced rehabilitation tools to empower individuals on their journey to recovery.

Industry 5.0 is bringing impressive technical and organizational advancements to the manufacturing sector. Technology and automation are now increasingly considered as partners to human workers, rather than as their substitutes, allowing scenarios of close collaboration that see humans as the core of any human-machine and human-robot interaction. In this framework, understanding and recognizing human activity during collaborative industrial operations is essential. In this talk, we provide an overview of our framework for sensitive and accurate psychophysiological monitoring of workload, fatigue, and stress of individuals interacting with a collaborative industrial robot (cobot). We particularly address challenges and lessons learned in using portable eye trackers for capturing pupillometry and other eye parameters (e.g., blinks) and chest straps for heart-rate variability (HRV) analysis.

Gait Analysis (GAn) is crucial for precision rehabilitation and clinical decisions. Traditionally conducted via tests and visual assessments, GAn measures spatio-temporal and kinematic parameters to detect gait abnormalities. Optical motion capture systems offer accuracy but are costly and lab-restricted. Marker-less motion analysis (MMA) represents a flexible, marker-free alternative, enabling faster, easier assessments and outdoor use, beneficial for sports biomechanics and daily monitoring of neuromotor disorders. MEMENTO develops a cost-effective, marker-less GAn software, harnessing pose estimation for motion detection using a simple camera, providing real-time feedback on walking performance. This innovation aims to improve the feasibility and usability of GAn in clinical settings, offering a unique training program for healthcare professionals globally in advanced gait assessment.

Parkinson's disease diagnosis using VitPose, a state-of-the-art pose estimation architecture, and through machine learning techniques. By employing a publicly accessible video dataset, we aim to accurately distinguish Parkinson's disease patients from healthy controls, ensuring transparency, reproducibility, and comparability in our methodology. A significant innovation of our study is integrating an explainability component, which provides clear insights into the decision-making process by identifying specific gait features indicative of Parkinson's disease. This integration enhances our research's robustness and clinical applicability and positions our approach as a valuable tool in clinical settings for diagnosing and monitoring Parkinson's disease.

Anastomotic leakage (AL) is the leaking of pathogenic, non-sterile gastrointestinal contents into the abdominal cavity of a patient. AL is one of the most dreaded complications after gastrointestinal surgery, with mortality rates of up to 27%. Existing diagnostic approaches for detecting anastomotic leaks exhibit limited sensitivity and specificity. Here, we introduce a naked eye-readable, electronic-free sensor placed in the extracorporeal drain bag for continuous monitoring and early detection of AL. As to date, there exists no satisfactory diagnostic method for AL detection, this work has the potential to significantly contribute to improved patient outcome through early, low-cost and reliable AL detection thereby reducing healthcare costs.

SafeLegs is a lower-limb exoskeleton prototype equipped with Deep Learning-based safety functions. First, we employed Reinforcement Learning for critical fault search in a Digital Twin of the exoskeleton. Secondly, by using Machine Learning optimisation, we drastically reduced amount of possible data access points from the system. From the optimised set of APs we trained the LSTM-based signal predictor. This model was quantised and deployed on the edge computer that controls the exoskeleton. As the result, the system is capable to successfully detect hardware faults before they can lead to a catastrophic failure while keeping the computational overhead as minimal as possible.

This document presents a new optimized integration of a compact gear reducer, specific designs of BLDC motor and magnetic absolute position sensor for high torque per volume demanding applications, and more particularly for an active exoskeleton dedicated to assisting disabled people or for a eWalk application.