Humanoid Robotics at Politecnico di Milano
Since 2000 we are developing a humanoid robot using a biomimetic approach.
The aim of our manipulation research is the development of a prototype of a human-like artificial arm. We base our design on the study of the natural limb, to emulate the kinematics and the strategy of control, replicating the spinal and cerebellum neural circuits. We started with the design of the hand, obtaining the Blackfingers prototype, actuated with McKibben muscles. Then we defined a simplified mechanical solution and produced Whitefingers.
G. Gini, M. Folgheraiter, "Blackfingers: an artificial hand that copies
human hand in structure, size, and function",Proc. IEEE Humanoids
M. Folgheraiter, G. Gini, M.
Perkowski," Adaptive Reflex Control for an Artificial Hand", 7th int
IFAC Symposium on robot control, Syroco 2003,
M. Folgheraiter, G. Gini, Human-like reflex control for an artificial hand, BioSystem Journal, ElsevierScience, 76, (1-3), August-October 2004, p 65-74.
Then we developed Maximum one, a full right arm.
Our robot arm differs in many aspects from other similar systems.
First, it has a full 3DOF joint for the shoulder, moved by five artificial muscles. Furthermore, we employed light materials and actuators. Finally, we attached the real prototype to a fully developed control system that simulates the motoneurones and cerebellar activities.
In order to experiment with the control architecture and compare its behavior to the human being, we also developed a kinematic and dynamic model of our arm prototype, and used it in simulations.
In particular we concentrated our work on testing a realistic artificial spinal circuit that demonstrated the capacity to replicate the human myotatic and inverse myotatic reflexes. Furthermore our experiments showed how it is possible to apply a bio-inspired control architecture also to an artificial system like our robotic arm. Surprisingly the system presented some behavior comparable with those of the natural limb.
G. Gini, "A Bio-inspired control system and a
VRML simulator for an autonomous humanoid arm",Proc. IEEE Humanoids 2003,October,
M. Folgheraiter, G. Gini, "MaximumOne: an anthropomorphic arm with bio-inspired control system", in Biomimetic Neural Learning for intelligent robots, S. Wernter, G. Palm, M. Elshaw (Eds), LNAI 3575, Springer, Berlin Heidelberg, 2005, pp 281-298.
To approach the programming of the arm through demonstration, we developed a special glove. The glove can detect the position of the fingers (angles of the phalanxes) and the force exerted by the user on the thumb and medium fingers. Moreover the glove can give contact feedback to the user on the fingertip (using a novel electro-stimulation device), and force feedback on the finger.
We are conducting experiments about a kind of imitation learning to teach the hand how to grasp objects.
We have developed a software environment where the user can see the objects being modified by touch. We are also developing new haptic devices for specific tasks.
The user haptic interface
The user command glove
The graphic environment
M. Folgheraiter, I. Baragiola, G.Gini, Teaching grasping to a humanoid hand as a generalization of human grasping data", in:Knowledge Exploration in Life Science Informatics (KELSI), Lecture notes in AI,Springer Verlag, November 2004. Pag 139-150.
M. Folgheraiter, G. Gini, D. Vercesi, "A Glove Interface with Tactile feeling display for Humanoid Robotics and Virtual Reality systems",Proc.International Conference ICINCO 2005, Barcellona 13-17 September, 2005.
M. Folgheraiter, G. Gini, D. Vercesi, "A new haptic device for applications in virtual reality and humanoid robotics", Integrated Computer-Aided Engineering, (IOSPress), Vol 13, N 3, 2006, pp 249-262.
For locomotion we are studying a mechanical design that exploit compliance. We have designed and built LARP.
The actuator system of LARP has a variable stiffness of the joints. The actuator in fact is a servomotor with a torsion spring and a damper.. In this way we can store energy in the joint to adsorb disturbances and we have an estimate of the external moment by measuring the deflection of the spring. Moreover we can obtain a real time control of the joint stiffness to adapt it to the external load to reduce, for instance, the inertial load.
We have developed a simulation using Adams and Matlab Simulink. The model has shown a good local stability to external disturbances, and is able to go back to the vertical position after being pushed.
LARP: the prototype
LARP: the foot
Scarfogliero, M. Folgheraiter, G. Gini, Advanced
Steps In Biped Robotics: Innovative Design And Intuitive Control Through
Spring-Damper Actuator Proc. IEEE Humanoids 2004,
U. Scarfogliero, M. Folgheraiter, G. Gini, "LARP, Biped Robot Conceived as Human Modelling", in Biomimetic Neural Learning for intelligent robots, S. Wernter, G. Palm, M. Elshaw (Eds), LNAI 3575, Springer, Heidelberg, 2005, pp 299 - 314.
U. Scarfogliero, M. Folgheraiter, "Human-oriented
biped robot design: insights into the development of a truly anthropomorphic
leg", Proc IEEE ICRA 07, 10-13 April 2007,
We are developing a robot head with stereovision and possible inclusion of other sensors.
The head for Maximum_one
A. Marchi, "Indoor Robot Navigation with Single Camera Vision",
Proc.Pattern Recognition in InformationSystems, PRIS,
We have developed a robot nose, he_knows, based on a matrix of sensors and devised to recognize and classify substances.
control panel and measuring chamber of he_knows
Our quadruped robots, ASGARD and WARUGADAR, have 12 dof and are actuated through electric servos. WARUGADAR uses piezoelectric sensors. Another quadruped, ULISSE, is pneumatically actuated.
leg and sensors of WARUGADAR
M. Folgheraiter, G. Gini, A. Nava, N. Mottola, "A bioinspired neural controller for a mobile robot", Proc 2006 IEEE International Conference on Robotics and Biomimetics (ROBIO 2006) Kunming, CHINA, December 17-20, 2006
P. Belluco, T. Ferrari, A new distribuite software architecture within the
Pyro environment for a quadruped robot, Proc IEEE CLAWAR, World Scientific
We are developing the applications needed to telecontrol the robot. They are interfaced to the user through our haptic device.
D. Librandi, "Teleprogramming a CRS robot through the internet",
Proc. IARP International worshop on Human Robot interfaces,
education, teleprogramming, telecontrol through the internet", Proc ISR2004, Paris (
Our study of the hand is also aimed at designing new prosthesis. We are working now on controlling them through EMG signals.
M. Folgheraiter, G. Gini,
M.Perkowski, M. Pivtoraiko, "Blackfingers: a
sophisticated hand prosthesis", Proc ICORRR 2003, International
Conference on rehabilitation robotics,
M. Arvetti, G. Gini, M. Folgheraiter Classification of EMG signals through wavelet analysis and neural networks for controlling an active hand prosthesis, Proc ICORR 2007, Noordwijk, The Netherlands, June 13-15, 2007.
The analysis of mio-electric signals is also the basis for controlling a patient rehabilitation using an exoscheleton.
M. Mulas, M. Folgheraiter, G. Gini, "An EMG-controlled Exoskeleton for Hand Rehabilitation ",Proc.International Conference ICORR 2005, Chicago.