As opposed to traditional robotic applications that require limited interaction and mobility, robots of the next generations will be required to physically interact with the environment and, on the long run, with humans. The dynamic interaction group activities aim at endowing humanoids with advanced action and physical interaction capabilities.
Our research is guided by two underpinning principles:
- studies on human motor control are relevant for building better performing robots;
- implementing humanoids contributes to better understanding the human motor control system.
Two different tools are at the basis of our research activities. On the one hand, iCub is the technological tool, which we contribute to develop. On the other hand, control theory is the theoretical toolset, which scientifically grounds the technological implementations. In the attempt to combine theoretical and technological development, our research activity produces sound theoretical results validated with dependable implementations on real robots. The main outcomes of our research activity are control principles, algorithms, technologies and implementations for compliant robot action and interaction. Our background is in engineering with strong competencies in physics, mathematics and computer science.
The dynamic interaction laboratory space hosts three research sub-groups: interaction modelling, estimation and control. Modelling deals with dynamic models of interaction, with the idea of simultaneously representing motion and forces. Estimation deals with the problem of accurate evaluation of those quantities that cannot be directly measured with sensors (e.g. internal forces). Control deals with the problem of controlling relevant interaction variables (e.g. exchanged forces). The group consists of about 10 researchers distributed on the three research topics. The laboratory includes one iCub, equipped with whole-body distributed tactile, inertial and force/torque sensing. Moreover, a Vicon system with force plates is available to conduct ground-truth inverse dynamics experiments in both humans and humanoids.
- FP7 CoDyCo (http://codyco.eu/): coordinator
- FP7 KOROIBOT (http://koroibot.eu/): scientific responsible for IIT
- FP7 RoboTAsk (http://www.robotask.eu): EU Marie Curie Actions with Francesca Stramandinoli
- H2020 SECURE (http://secure-robots.eu): Marie Skłodowska-Curie Innovative Training Networks with Giorgio Metta
- H2020 PACE (http://itn-pace.eu): Marie Skłodowska-Curie Innovative Training Networks with Gabriel Baud-Bovy
Whole-body software development
The dynamic interaction control laboratory collaborates with Nikos Tsagarakis at the Advanced Robotics Department and Lorenzo Natale at the iCub facility on the topic of software development of whole-body motion control. Within the YARP software framework, they are developing software tools for modelling and controlling articulated rigid bodies.
The dynamic interaction control laboratory collaborates with Lorenzo Natale and Giorgio Metta at the iCub facility on the development of the iCub humanoid. The focus is on aspects related to control including actuation, sensors, software, hardware and communication. The dynamic interaction control laboratory collaborates also with Luca Fiorio and Giorgio Metta at the iCub facility and with Nikos Tsagarakis on the topic of compliant actuation. Within this collaboration, the focus of my research activity is on building novel compliant actuators, which exploit a property of human muscles, nominally the ability to reject disturbances without explicitly resorting on feedback.
Analysing human biomechanics
The dynamic interaction control laboratory collaborates with with Bastien Berret at the University Paris-Sud on the topic of human biomechanical modelling. This research activity aims at understanding how the biomechanical properties of human muscles (e.g. variable compliance) influence and, in a sense simplify, the execution of complex control tasks.
Human force-motion tracking
The dynamic interaction control laboratory collaborates with Xsens Technologies B.V. on the topic of human whole-body dynamics estimation. This research focuses on building wearable devices for on-line estimating human whole-body dynamics from multiple sensor modalities (e.g. gyroscopes, accelerometers, force-platform and EMG).
The dynamic interaction control laboratory collaborates with Alessandro Chiuso at the Università di Padova and with Lorenzo Rosasco at the Università di Genova on the topic of whole-body dynamics identification. This research aims at building dynamic models of articulated rigid body structures preserving properties (e.g. system passivity) which are fundamental for use of the identified models in synthesizing whole-body motion controllers.