|Participant organisation name||Short name||Participant organisation Country||Principal roles|
|Fondazione Istituto Italiano di Tecnologia (coordinator)||IIT||ITA||Francesco Nori (Coordinator)
Daniele Pucci (PI)
|Istituto Nazionale Assicurazione contro gli Infortuni sul Lavoro||INAIL||ITA||Sergio Iavicoli|
|Institut national de recherche en informatique et en automatique||INRIA||FRA||Serena Ivaldi (PI)|
|Institut Jožef Stefan||JSI||SVN||Jan Babic (PI)|
|Deutsches Zentrum für Luft- und Raumfahrt||DLR||DE||Freek Stulp (PI)|
|Xsens Technologies||XSENS||NL||Giovanni Bellusci|
|IMK automotive GmbH (SME)||IMK||DE||Lars Fritzsche|
|Otto Bock HealthCare GmbH||OBGH||DE||Graimann Bernhard|
|AnyBody Technology A/S (SME)||ABT||DK||Michael Damsgaard|
The Fondazione Istituto Italiano di Tecnologia brings in the consortium its expertise in tactile technologies, humanoid technologies, whole-body dynamics estimations in human and humanoids. Francesco Nori has a longstanding experience in EU projects and brings in ANDY his expertise in coordinating EU projects of similar structure (CoDyCo).
L’Istituto Nazionale per l'Assicurazione contro gli Infortuni sul Lavoro e le malattie professionali
Inail, the Italian National Institute for Insurance against Accidents at Work, is a public non-profit entity safeguarding workers against physical injuries and occupational diseases.
Director of INAIL Department of Occupational and Environmental Medicine, Epidemiology and Hygiene
INRIA contribution will be in the acquisiton of the ANDYDATASET, then in the development of modeling, learning and control strategies for human-robot physical interaction and machine learning. The first contribution will be in the acquisition of experimental datasets of humans interacting with robots (ANDYDATASET). We will organize the acquisition in research labs and in end-users sites, thanks to the collaboration of the end-users of the advisory board. The second contribution will be in the application and development of machine learning strategies for controlling the robot during pHRI. The third contribution will be in the realization of the scenario of human-humanoid collaborative assembly with the iCub humanoid. Finally, INRIA has several managing responsabilities, in terms of ethics, data management, exploitation management and relations with the end-user advisory board. For these activities, the PI and the heads from the Department of Industrial Relations of INRIA will cooperate with the project governance to ensure the successful exploitation of the project outcomes.
The role of JSI will be in development of adequate dynamic models of the human body in tight physical interaction with an industrial collaborative robot for Scenario 1, an exoskeleton for Scenario 2 and a humanoid robot for Scenario 3. The common denominator of all three scenarios is the extension of human body schema and the consideration of the interacting robotic system as a tool operated by the human. Special attention will be devoted to understand the dynamic properties of the multi-contact interaction between the human and the robotic system. To facilitate control and machine learning in all three scenarios, modelling will be leveraged by the use of optimization methods to develop optimality principles underlying human motion in collaboration/interaction with the robotic systems.
The main contribution of DLR revolves around learning predictive models through machine learning (ML). In particular, DLR will conduct basic research on all validation scenarios through contributions on Objectives 2 and 3. The ANDYSUIT developed as a result of Objective 1 will provide an unprecedented wealth of data recorded live from a human subject, namely contact forces, joint torques and muscle activations detected through surface electromyography, tactile and pressure sensors, stretch sensing and so on. Using these data, within Objective 2, DLR will develop innovative methodologies to classify human motion and to diagnose human musculoskeletal disorders; classification will be enforced in real time during physical interaction (e.g., pulling, pushing, peg inserting, etc.). Off-line regression techniques will be used to model human-joint torques while performing interaction tasks. DLR will also provide the system and software for realizing the use cases in Scenario 1, see also “Significant infrastructure” below, and play an assistive role in applying reinforcement learning techniques to synthesise efficient controllers for human-robot physical interaction. Lastly, given the DLR’s expertise in the field, the institute will assist with the acquisition, calibration and testing of the EMG sensors; provide knowledge when recording data from them; and actively employ them on-line in the application Scenarios.
Xsens Technologies BV coordinates the ANDY activities for developing a wearable force and motion tracking technology (ANDYSUIT and Objective 1). XSENS brings into the consortium its experience in the development and commercialization of a wearable motion tracking system, the MVN Biomech (https://www.xsens.com/products/mvn-biomech/). This technology will be enhanced with additional wearable sensors (e.g. the Xsens Force Shoes) to make it suitable for simultaneous force and motion tracking. This technology will impact on all the ANDY scenarios (Scenarios 1, 2 and 3) and therefore collaborations are foreseen with the WP5 leaders. Within WP1, XSENS will collaborate with IIT, which has a background in on-line whole-body dynamics estimation.
IMK will bring two key contributions in the project: long practical experience in ergonomic risk assessment and outstanding expertise in digital human modelling and human work simulation. The main role of IMK is to develop a software prototype that is able to simulate human-robot collaboration based on sensor data collected from ANDYSUIT. The software will also be able to estimate ergonomic risks and provide indications for ergonomic improvements. Current methods for ergonomic risk assessment, which are mainly based on observation data, will be advanced by taking into account the objectively measured force and movement data provided by the ANDYSUIT. The new software prototype will serve as a foundation for significantly extending the capabilities of the existing “EMA” (Editor for Manual Work Activities) software. It can then be used for conducting virtual tests for human-robot collaboration scenarios in terms of safety, ergonomics and efficiency. Additionally, IMK will support the setup, execution and evaluation of pilot cases and also support the dissemination and exploitation of project results by using its large network of research partners and industrial clients.
Otto Bock will take care of the following activities:
- define the and contribute to the requirements of technologies, prototypes and evaluation scenarios of this project;
- play a main role in the Scenario 2 (support for manual labor tasks – upper extremity);
- adapt an already existing orthotic device for upper limb support in car assembly and develop it further to an actuated exoskeleton
- co-develop the interfaces to the sensor-system (ANDYSUIT);
- develop together with IIT the control mechanisms for the exoskeleton;
- OBGH will participate in the evaluation of the Scenario 2
AnyBody Technology A/S (SME) contributes to the project with its expertise in detailed musculoskeletal modeling, here-under modeling of man-machine interfaces. This plays a fundamental role in the dynamic modeling activities; not only to develop the models but also to find the best parameters and tuning approaches for the models for a given subject. Furthermore, ABT expertise will be exploited in the development of the ANDYDATASET dataset and in the development of the estimation technologies. Musculoskeletal models will be one of the tools to estimate properties that cannot be measured directly or accurately. Detailed musculoskeletal models will also be a tool to accurately estimate parameters that are being approximated during the real-time applications and as such, they can be used while verifying the accuracy of these applications.
A primary focus of ABT will be to bring as much of the detailed musculoskeletal model information into the real-time dynamics models as possible and to bring as much of efficiency as possible back into the detailed musculoskeletal models for independent usage, for instance as part of detailed post-processing ergonomic analysis of scenarios such as the three scenarios of the project. Enhanced ergonomic assessment models based on detailed musculoskeletal model data are also a focus area of ABT.