Markerless visual servoing on unknown objects for humanoid robot platforms iCub HumanoidRobot 2017-11-10 | This video shows a new framework for markerless visual servoing on unknown objects in action. The pipeline consists of four main parts:1) a least- squares minimization problem is formulated to find the volume of the object graspable by the robot’s hand using its stereo vision;2) a recursive Bayesian filtering technique, based on Sequential Monte Carlo (SMC) filtering, estimates the 6D pose (position and orientation) of the robot’s end-effector without the use of markers;3) a nonlinear constrained optimization problem is formulated to compute the desired graspable pose about the object;4) an image-based visual servo control commands the robot’s end-effector toward the desired pose.The method is described in the following preprint arXiv paper:C. Fantacci, G. Vezzani, U. Pattacini, V. Tikhanoff and L. Natale, "Markerless visual servoing on unknown objects for humanoid robot platforms", arXiv preprint arXiv:1710.04465, 2017.
Learn Fast, Segment Well: Fast Object Segmentation Learning on the iCub Robot. iCub HumanoidRobot 2022-06-28 | This work demonstrates a pipeline for fast instance segmentation learning designed for robotic applications where data come in stream. The main contributions are a modified version of Mask-RCNN and training protocol specifically designed to allow feature extraction to happen on-line during data acquisition. The code to reproduce the experiments is publicly available on GitHub (github.com/hsp-iit/online-detection). Details of the work are reported in the paper: Ceola, F., Maiettini, E., Pasquale, G., Manti G., Rosasco, L. and Natale, L., "Learn Fast, Segment Well: Fast Object Segmentation Learning on the iCub Robot", IEEE Transactions on Robotics, 2022.
Redball++ iCub HumanoidRobot 2022-03-07 | This video shows the integration of the pose tracking and grasping with superquadric functions on the iCub humanoid robot.Details of the components used in this work: Piga, N., Onyshchuk, Y., Pasquale, G., Pattacini, U., and Natale, L., ROFT: Real-time Optical Flow-aided 6D Object Pose and Velocity Tracking, IEEE Robotics & Automation Magazine, vol. 7, no. 1, pp. 159-166, 2022 ieeexplore.ieee.org/document/9568706.Nguyen, P. D. H., Bottarel, F., Pattacini, U., Hoffmann, M., Natale, L., and Metta, G., Merging Physical and Social Interaction for Effective Human-Robot Collaboration, in Proc. IEEE-RAS International Conference on Humanoid Robots, Beijing, China, 2018, pp. 1-9.
Active Perception for Ambiguous Objects Classification iCub HumanoidRobot 2021-09-24 | Some objects cannot be discriminated by looking at a single view. In this video we demonstrate an active perception system that is able to i) determine whether a given view is insufficient to discriminate an object, and ii) select the next best view for discrimination. Details of this paper are available in:Safronov, Piga, Colledanchise & Natale, Active Perception for Ambiguous Objects Classification, IROS 2021. Full text: arxiv.org/abs/2108.00737
A Differentiable Extended Kalman Filter for Object Tracking Under Sliding Regime iCub HumanoidRobot 2021-09-02 | In this work, we propose a differentiable Extended Kalman filter that can be trained to track the position and the velocity of an object under translational sliding regime from tactile observations alone. Details of this work can be found in the paper:A Differentiable Extended Kalman Filter for Object Tracking Under Sliding Regime, Piga, N.A, Pattacini, U. and Natale, L., Frontiers in robotics and AI, 2021.Full text: frontiersin.org/articles/10.3389/frobt.2021.686447/full 1 Introduction
Weakly-Supervised Object Detection Learning through Human-Robot Interaction iCub HumanoidRobot 2021-07-12 | This video demonstrates a system for efficiently training an object detection system on a humanoid robot. The proposed system allows to iteratively adapt an object detection model to novel contexts, by exploiting: (i) a teacher-learner pipeline, (ii) weakly supervised learning techniques to reduce the human labeling effort and (iii) an on-line learning approach for fast model re-training.Details of this work can be found in the following paper: Maiettini E., Tikhanoff V., Natale L., Weakly-Supervised Object Detection Learning through Human-Robot Interaction, IEEE-RAS International Conference on Humanoid Robots, 2021
On the Emergence of Whole-body Strategies from Humanoid Robot Push-recovery Learning iCub HumanoidRobot 2021-07-12 | Diego Ferigo*, Raffaello Camoriano*, Paolo Maria Viceconte, Daniele Calandriello, Silvio Traversaro, Lorenzo Rosasco, Daniele Pucci, "On the Emergence of Whole-body Strategies from Humanoid Robot Push-recovery Learning", IEEE Robotics Automation Letters (RA-L) 2021 and IEEE Humanoids 2020* Authors equally contributed to this paperProject page and supplementary videos: dic-iit.github.io/emergence-push-recovery-icubPre-print: arxiv.org/abs/2104.14534AbstractBalancing and push-recovery are essential capabilities enabling humanoid robots to solve complex locomotion tasks. In this context, classical control systems tend to be based on simplified physical models and hard-coded strategies. Although successful in specific scenarios, this approach requires demanding tuning of parameters and switching logic between specifically-designed controllers for handling more general perturbations. We apply model-free Deep Reinforcement Learning for training a general and robust humanoid push-recovery policy in a simulation environment. Our method targets high-dimensional whole-body humanoid control and is validated on the iCub humanoid. Reward components incorporating expert knowledge on humanoid control enable fast learning of several robust behaviors by the same policy, spanning the entire body. We validate our method with extensive quantitative analyses in simulation, including out-of-sample tasks which demonstrate policy robustness and generalization, both key requirements towards real-world robot deployment.Authors detailsDiego Ferigo*: Dynamic Interaction Control (DIC), Istituto Italiano di Tecnologia, ITA; ML and Optimisation, University of Manchester, UKRaffaello Camoriano*: Laboratory for Computational and Statistical Learning (IIT@MIT), Istituto Italiano di Tecnologia, ITAPaolo Maria Viceconte: Dynamic Interaction Control (DIC), Istituto Italiano di Tecnologia, ITA; DIAG, Sapienza Università di Roma, ITADaniele Calandriello: Laboratory for Computational and Statistical Learning (IIT@MIT), Istituto Italiano di Tecnologia, ITASilvio Traversaro: DIC, Istituto Italiano di Tecnologia, ITALorenzo Rosasco: Machine Learning Genoa Center (MaLGa) & DIBRIS, University of Genoa, Italy, ITA; Laboratory for Computational and Statistical Learning (IIT@MIT), Istituto Italiano di Tecnologia, ITA; Center for Brains, Minds and Machines, MIT, USADaniele Pucci: Dynamic Interaction Control (DIC), Istituto Italiano di Tecnologia, ITA
Shared Control of Robot-Robot Collaborative Lifting with Agent Postural and Force Ergonomic Optimiza iCub HumanoidRobot 2021-05-31 | Humans show specialized strategies for efficient collaboration. Transferring similar strategies to humanoid robots can improve their capability to interact with other agents, leading the way to complex collaborative scenarios with multiple agents acting on a shared environment. In this paper we present a control framework for robot-robot collaborative lifting. The proposed shared controller takes into account the joint action of both the robots thanks to a centralized controller that communicates with them, and solves the whole-system optimization. Efficient collaboration is ensured by taking into account the ergonomic requirements of the robots through the optimization of posture and contact forces. The framework is validated in an experimental scenario with two iCub humanoid robots performing different payload lifting sequences.
In-Situ Translational Hand-Eye Calibration of Laser Profile Sensors Using Arbitrary Objects iCub HumanoidRobot 2021-05-31 | Preprint at: arxiv.org/pdf/2103.11981.pdf Abstract. Hand-eye calibration of laser profile sensors is the process of extracting the homogeneous transformation between the laser profile sensor frame and the end-effector frame of a robot in order to express the data extracted by the sensor in the robot’s global coordinate system. For laser profile scanners this is a challenging procedure, as they provide data only in two dimensions and state-of-the-art calibration procedures require the use of specialised calibration targets. This paper presents a novel method to extract the translation-part of the hand-eye calibration matrix with rotation-part known a priori in a target agnostic way. Our methodology is applicable to any 2D image or 3D object as a calibration target and can also be performed in situ in the final application. The method is experimentally validated on a real robot-sensor setup with 2D and 3D targets.
Modeling of Visco-Elastic Environments for Humanoid Robot Motion Control iCub HumanoidRobot 2021-05-31 | This manuscript presents a model of compliant contacts for time-critical humanoid robot motion control. The proposed model considers the environment as a continuum of spring-damper systems, which allows us to compute the equivalent contact force and torque that the environment exerts on the contact surface. We show that the proposed model extends the linear and rotational springs and dampers -- classically used to characterize soft terrains -- to the case of large contact surface orientations. The contact model is then used for the real-time whole-body control of humanoid robots walking on visco-elastic environments. The overall approach is validated by simulating walking motions of the iCub humanoid robot. Furthermore, the paper compares the proposed whole-body control strategy and state of the art approaches. In this respect, we investigate the terrain compliance that makes the classical approaches assuming rigid contacts fail. We finally analyze the robustness of the presented control design with respect to non-parametric uncertainty in the contact-model.
DILIGENT-KIO: A Proprioceptive Base Estimator for Humanoid Robots Using EKF on Matrix Lie Groups iCub HumanoidRobot 2021-05-31 | This paper presents a contact-aided inertial-kinematic floating base estimation for humanoid robots considering an evolution of the state and observations over matrix Lie groups. This is achieved through the application of a geometrically meaningful estimator which is characterized by concentrated Gaussian distributions. The configuration of a floating base system like a humanoid robot usually requires the knowledge of an additional six degrees of freedom which describes its base position-and-orientation. This quantity usually cannot be measured and needs to be estimated. A matrix Lie group, encapsulating the position-and-orientation and linear velocity of the base link, feet positions-and-orientations and Inertial Measurement Units' biases, is used to represent the state while relative positions-and-orientations of contact feet from forward kinematics are used as observations. The proposed estimator exhibits fast convergence for large initialization errors owing to choice of uncertainty parametrization. An experimental validation is done on the iCub humanoid platform.
Online segmentation iCub HumanoidRobot 2021-05-21 | In this work, we propose a novel architecture for online learning of object segmentation that provides comparable performance to SOA in a fraction of the training time. All details are reported in the paper:Ceola, F., Maiettini, E., Pasquale, G., Rosasco, L., and Natale, L., Fast Object Segmentation Learning with Kernel-based Methods for Robotics, in Proc. IEEE-RAS International Conference on Robotics and Automation, 2021.Link to pdf: arxiv.org/pdf/2011.12805.pdf
MaskUKF: 6D Object Pose and Velocity Tracking iCub HumanoidRobot 2021-03-23 | Tracking the 6D pose and velocity of objects represents a fundamental requirement for modern robotics manipulation tasks. This paper proposes a 6D object pose tracking algorithm, called MaskUKF, that combines deep object segmentation networks and depth information with a serial Unscented Kalman Filter to track the pose and the velocity of an object in real-time. MaskUKF achieves and in most cases surpasses state-of-the-art performance on the YCB-Video pose estimation benchmark without the need for expensive ground truth pose annotations at training time. Closed loop control experiments on the iCub humanoid platform in simulation show that joint pose and velocity tracking helps achieving higher precision and reliability than with one-shot deep pose estimation networks.Technical description of the material described in the video can be found in the following paper:Piga, A.N., Bottarel, F., Fantacci, C., Vezzani G., Pattacini, U. and Natale, L. MaskUKF: An Instance Segmentation Aided Unscented Kalman Filter for 6D Object Pose and Velocity Tracking, Frontiers in Robotics and AI, 2021.frontiersin.org/articles/10.3389/frobt.2021.594583/full
Model-Based Real-Time Motion Tracking using Dynamical Inverse Kinematics iCub HumanoidRobot 2021-01-22 | This video presents the paper published to MDPI Algorithms entitled "Model-Based Real-Time Motion Tracking using Inverse Kinematics" mdpi.com/1999-4893/13/10/266/htm. It represents the latest achievement of the Dynamic Interaction Control lab https://dic.iit.it/ in the whole body motion tracking of human beings.
Jerk Control of Floating Base Systems With Contact-Stable Parameterized Force Feedback iCub HumanoidRobot 2020-12-17 | This video presents the paper published to IEEE Transaction on Robotics entitled "Jerk Control of Floating Base Systems With Contact-Stable Parameterized Force Feedback" ieeexplore.ieee.org/document/9237133. It represents the latest achievement of the Dynamic Interaction Control lab https://dic.iit.it/ in the whole body control of humanoid robots.
Formal verification and runtime monitoring of behavior trees (SCOPE project, final release) iCub HumanoidRobot 2020-12-04 | SCOPE is one of the Integrated Technical Project ITP funded by the second open call within the RobMoSys EU project. SCOPE provide tools that analyze and derive properties of a task by composing the properties that describe its skills and the environment, and, at runtime, ensure the correct execution of a task by monitoring it and propagating anomalies detected at the level of the skills. This video shows the final (third) release of the validation scenario. More information: scope-robmosys.github.io/.
Formal verification and runtime monitoring of behavior trees (SCOPE project, release 2) iCub HumanoidRobot 2020-11-20 | SCOPE is one of the Integrated Technical Project ITP funded by the second open call within the RobMoSys EU project. SCOPE provide tools that analyze and derive properties of a task by composing the properties that describe its skills and the environment, and, at runtime, ensure the correct execution of a task by monitoring it and propagating anomalies detected at the level of the skills. This video shows the second release of the validation scenario. More information: scope-robmosys.github.io/.
Modeling, Identification and Control of Model Jet Engines for Jet Powered Robotics iCub HumanoidRobot 2020-06-04 | This video presents the paper entitled "Modeling, Identification and Control of Model Jet Engines for Jet Powered Robotics" published in IEEE Robotics and Automation Letters (Volume: 5 , Issue: 2 , April 2020 ) Page(s): 2070 - 2077. Preprint at arxiv.org/pdf/1909.13296.pdf. Lab website: https://dic.iit.it
iRonCub Flight Simulation in a Disaster Scenario iCub HumanoidRobot 2020-06-04 | This video shows the latest results on Aerial Humanoid Robotics obtained by the Dynamic Interaction Control Lab https://dic.iit.it/ at the Italian Institute of Technology. The video simulates robot and jet dynamics, where the latter uses the results obtained in the paper "Modeling, Identification and Control of Model Jet Engines for Jet Powered Robotics" published in IEEE Robotics and Automation Letters (Volume: 5 , Issue: 2 , April 2020 ) Page(s): 2070 - 2077 (preprint at arxiv.org/pdf/1909.13296.pdf, video at youtube.com/watch?v=DJv5aXLhQIM). Furthermore, the developed simulator integrates sound simulation, which originates from the real experiments performed for jet identification and control purposes. Finally, the control algorithms ensure vertical take-off and landing, with orientation control and flight information.
Act, Perceive and Plan in Belief Space for Robot Localization iCub HumanoidRobot 2020-05-19 | We propose an interleaved acting and planning technique to rapidly reduce the uncertainty of the estimated robot’s pose by perceiving relevant information from the environment, as recognizing an object or asking someone for a direction.This video demonstrates work published in:M. Colledanchise, D. Malafronte and L. Natale, Act, Perceive, and Plan in Belief Space for Robot Localization, ICRA 2020. Link to PDF: arxiv.org/pdf/2002.08124
Grasping and Navigation with Behavior Trees CARVE Scenario 1 iCub HumanoidRobot 2020-05-19 | This video shows the integration between SmartSoft and YARP, and the Behavior Tree engine developed in the CARVE project.It demonstrates the execution of a Behavior Tree for "fetching an object". The BT controls the robot to navigate to the kitchen, approach the table and locate the bottle using its visual system. Finally, the BT sends the commands to compute a valid grasping pose for the hand, reposition the robot, and lift the object from the table.Acknowledgements:This work was carried out in the context of the CARVE project, which has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No 732410, in the form of financial support to third parties of the RobMoSys project.
Grasping, Navigation and Human follower with Behavior Trees - CARVE Scenario 2 iCub HumanoidRobot 2020-05-19 | This video extends Scenario 1 behavior tree and demonstration (youtu.be/qIFLg4F-06w), with a person follower. In case the room is not known to the robot, he can asks the user to show where the room is.Scenario 2 demonstrates integration of Behavior Tree engine in YARP and SmartSoft.Acknowledgements:This work was carried out in the context of the CARVE project, which has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No 732410, in the form of financial support to third parties of the RobMoSys project.
Pouring a drink iCub HumanoidRobot 2020-05-19 | This video shows the integration between SmartSoft and YARP, and the Behavior Tree engine developed in the CARVE project. This video demonstrates Scenario 3: pouring a drink.
Whole-Body Geometric Retargeting for Humanoid Robots iCub HumanoidRobot 2019-10-04 | This video shows the latest results of the Dynamic Interaction Control lab https://dic.iit.it in the whole teleoperation of humanoid robots. Paper in the proceedings of 2019 IEEE-RAS International Conference on Humanoid Robots. Preprint at arxiv.org/abs/1909.10080
Online DCM Trajectory Generation for Push Recovery of Torque-Controlled Humanoid Robots iCub HumanoidRobot 2019-09-24 | This video shows the latest results of the Dynamic Interaction Control lab https://dic.iit.it in the whole body walking of humanoid robots. Paper in the proceedings of 2019 IEEE-RAS International Conference on Humanoid Robots. Preprint at arxiv.org/abs/1909.10403AbstractWe present a computationally efficient method for online planning of bipedal walking trajectories with push recovery. In particular, the proposed methodology fits control architectures where the Divergent-Component-of-Motion(DCM) is planned beforehand, and adds a step adapter to adjust the planned trajectories and achieve push recovery. Assuming that the robot is in a single support state, the step adapter generates new positions and timings for the next step. The step adapter is active in single support phases only, but the proposed torque-control architecture considers double support phases too. The key idea for the design of the step adapter is to impose both initial and final DCM step values using an exponential interpolation of the time varying ZMP trajectory.This allows us to cast the push recovery problem as a QuadraticProgramming (QP) one, and to solve it online with state-of-the-art optimisers. The overall approach is validated with simulations of the torque-controlled 33 kg humanoid robot iCub. Results show that the proposed strategy prevents the humanoid robot from falling while walking at 0.28 m/s and pushed with external forces up to 150 Newton for 0.05 seconds.
Automatic Creation of Large Scale Object Databases from Web Resources: A Case Study in Robot Vision iCub HumanoidRobot 2019-09-10 | This video demonstrates a method for training a deep neural network online, using task specific datasets automatically created by exploiting the Web as a source of information.Details of this work are described in:Dario Molinari, Giulia Pasquale, Lorenzo Natale and Barbara Caputo, Automatic Creation of Large Scale Object Databases from Web Resources: A Case Study in Robot Vision, ICIAP 2019 (link.springer.com/chapter/10.1007/978-3-030-30645-8_45).Video credits: Giulia Pasquale, Dario Molinari, Vadim Tikhanoff, Lorenzo Natale and Barbara Caputo, Istituto Italiano di Tecnologia.
A Benchmarking of DCM Based Architectures for Walking Humanoid Robots iCub HumanoidRobot 2019-04-02 | Paper published in IEEE Humanoids 2018 ieeexplore.ieee.org/document/8625025Selected as one of the outstanding papers of the IEEE Humanoids 2018Arxiv link arxiv.org/abs/1809.02167Dynamic Interaction Control https://www.dic.iit.it/AbstractThis paper contributes towards the development and comparison of Divergent-Component-of-Motion (DCM) based control architectures for humanoid robot locomotion. More precisely, we present and compare several DCM based implementations of a three layer control architecture. From top to bottom, these three layers are here called: trajectory optimization, simplified model control, and whole-body QP control. All layers use the DCM concept to generate references for the layer below. For the simplified model control layer, we present and compare both instantaneous and Receding Horizon Control controllers. For the whole-body QP control layer, we present and compare controllers for position and velocity control robots. Experimental results are carried out on the one-meter tall iCub humanoid robot. We show which implementation of the above control architecture allows the robot to achieve a walking velocity of 0.41 meters per second.
Dynamic Interaction Control labs 2018 Year In Review iCub HumanoidRobot 2019-01-02 | This video reviews the research results of the Dynamic Interaction Control lab obtained in 2018. Visit the website https://www.dic.iit.it/ to know more about the DIC lab research directions and results
Speeded up training of object detection iCub HumanoidRobot 2018-11-20 | Latest results on visual perception with the R1 robot: online training of object detection network. This video demonstrates an implementation of the work described in this paper (presented at IROS 2018, madrid):Maiettini, E., Pasquale, G., Rosasco, L., and Natale, L., Speeding-up Object Detection Training for Robotics with FALKON, in Proc. IEEE/RSJ International Conference on Intelligent Robots and Systems, Madrid, Spain 2018 (preprint available: arxiv.org/abs/1803.08740).
Continuous listening, distant spoken command recognizer for human-robot interaction iCub HumanoidRobot 2018-10-31 | This is a first demonstrator of an automatic command recognition system for R1. Its main features are:- Continuous listening. The system is always active and ready to recognize commands addressed to R1. It does not require any keywords to activate command recognition (e.g., “Ehi R1”, “OK R1”)- Distant speech recognition. The system recognizes the commands with high accuracy even when the speaker is free to move within a 3m radius.The system is based on deep neural networks and on our “articulatory” approach to automatic speech recognition.
iCub teleoperated walking and manipulation iCub HumanoidRobot 2018-09-12 | First results on telexistence of the Dynamic Interaction Control https://www.dic.iit.it/Cite this contribution - Teleoperation: Mohamed Elobaid, Yue Hu, Jan Babic, and Daniele Pucci. “Telexistence and Teleoperation for Walking Humanoid Robots”. In: 2019 Intelligent Systems Conference (IntelliSys), London, 2017, in press.- Walking: G. Romualdi, S. Dafarra, Y. Hu, and D. Pucci. “A Benchmarking of DCM Based Architectures for Position and Velocity Controlled Walking of Humanoid Robots”. In: 2018 IEEE-RAS 18th International Conference on Humanoid Robots (Humanoids). 2018, pp. 1–9. doi: 10.1109/ HUMANOIDS.2018.8625025., online at arxiv.org/abs/1809.02167This video shows the latest results achieved by the Dynamic Interaction Control Lab at the Italian Institute of Technology on teleoperated walking and manipulation for humanoid robots.We have integrated the iCub walking algorithms with a new teleoperation system, thus allowing a human being to teleoperate the robot during locomotion and manipulation tasks.
Task-based Control of an Underactuated Flying Humanoid Robot iCub HumanoidRobot 2018-07-28 | This video shows the simulations results presented in the paper "Task-based Control of an Underactuated Flying Humanoid Robot" submitted for possible publication to IEEE Humanoids 2018. Differently from goo.gl/R52aXs we here achieve stability and convergence of not only the robot momentum, but also its center-of-mass and pelvis orientation. This allows us to better control aggressive robot flight manoeuvres. Also, simulations are now performed in the Gazebo environment with more realistic jet turbines models.
Compact real-time avoidance on a humanoid robot for human-robot interaction iCub HumanoidRobot 2017-12-13 | This video shows the ability of iCub to work safely in sharing environment with human.Taking inspiration from peripersonal space representations in humans, we present a framework on the iCub humanoid robot that dynamically maintains such a protective safety zone, composed of the following main components:- An architecture for human keypoints estimation in 3D- An adaptive peripersonal space representation- A controller dynamically incorportating human keypoints as obstacles into a reaching taskThe video is based on the work presented in:Dong Hai Phuong Nguyen, Matej Hoffmann, Alessandro Roncone, Ugo Pattacini, and Giorgio Metta. 2018. Compact Real-time Avoidance on a Humanoid Robot for Human-robot Interaction. In HRI ’18: 2018 ACM/IEEE International Conference on Human-Robot Interaction, March 5–8, 2018, Chicago, IL, USA. ACM, New York, NY, USA, 9 pages. pdf: arxiv.org/abs/1801.05671 doi.org/10.1145/3171221.3171245
Visual end-effector tracking using a 3D model-aided particle filter for humanoid robot platforms iCub HumanoidRobot 2017-08-03 | This video demonstrates recursive markerless estimation of a robot’s end-effector using visual observations from its cameras. The problem is formulated into the Bayesian framework and addressed using Sequential Monte Carlo (SMC) filtering. We demonstrate that the tracking is robust to clutter, allows compensating for errors in the robot kinematics and servoing the arm in closed loop using vision.The method is described in the following paper:C. Fantacci, U. Pattacini, V. Tikhanoff and L. Natale, "Visual end-effector tracking using a 3D model-aided particle filter for humanoid robot platforms", IEEE/RSJ International Conference on Intelligent Robots and Systems, Vancouver, BC, Canada, September 24-28, 2017.
The robot homunculus: learning of artificial skin representation inspired by the brain iCub HumanoidRobot 2017-07-14 | Leveraging on the artificial skin of the iCub humanoid, researchers at the iCub Facility of the Istituto Italiano di Tecnologia, together with the Czech Technical University in Prague and Comenius University in Bratislava, studied how can the robot develop a representation of his body surface only from being touched by other people. Human and monkey brains contain distorted body maps of the body - the so-called "homunculi", or "little men". A new algorithm was used that allowed the iCub to develop the "robot homunculus" for its brain -- a map of its skin surface similar to the one in biological brains. In this way, the engineers do not need to calibrate and program the skin model anymore and the iCub can use the homunuculus he learned autonomously. Furthermore, the algorithm can cope with changes to the body: when one body part stops functioning, the robot "brain" reallocates this territory to other body parts. With the artificial skin as a key enabling technology, the representation developed here is being integrated into new control algorithms that improve the safety of human-robot interaction and pave the way for robots leaving the factories and entering our homes.Full details can be found in: Hoffmann, M.; Straka, Z.; Farkas, I.; Vavrecka, M. & Metta, G. (2017), 'Robotic homunculus: Learning of artificial skin representation in a humanoid robot motivated by primary somatosensory cortex', IEEE Transactions on Cognitive and Developmental Systems. doi.org/10.1109/TCDS.2017.2649225
A Grasping Approach Based on Superquadric Models iCub HumanoidRobot 2017-05-22 | This video demonstrates grasping of objects using superquadric models. The method is described in the following paper:Vezzani, G., Pattacini, U., and Natale, L., "A Grasping Approach Based on Superquadric Models", in IEEE International Conference on Robotics and Automation, Singapore, 2017
Robust Visual Tracking with a Freely moving Event Camera iCub HumanoidRobot 2017-03-01 | The iCub follows the moving target using event-based cameras. A novel event-based particle filter tracks the ball position at over 200Hz.
Hierarchical Grasp Controller using Tactile Feedback iCub HumanoidRobot 2016-10-18 | iCub uses tactile feedback to control and improve object grip.Details in: Hierarchical Grasp Controller using Tactile Feedback, M. Regoli, U. Pattacini, Metta, G. and Natale, L., Humanoids 2016.
Pressure control using tactile sensors iCub HumanoidRobot 2016-08-30 | This video shows control of pressure using the tactile sensors on the fingertips of the iCub. The robot exerts increasing pressure on the object. Credits: Massimo Regoli, Ugo Pattacini & Lorenzo NataleThe tactile sensors on the fingertips are described here: Jamali, N., Maggiali, M., Giovannini, F., Metta, G., and Natale, L., A New Design of a Fingertip for the iCub Hand, in Proc. IEEE/RSJ International Conference on Intelligent Robots and Systems, Hamburg, Germany, 2015, pp. 1799-1805
iCub performing highly dynamic Tai Chi while interacting with humans iCub HumanoidRobot 2016-07-29 | Cite this contribution:D. Pucci and F. Romano and S. Traversaro and F. Nori; "Highly dynamic balancing via force control" 2016 IEEE-RAS 16th International Conference on Humanoid Robots (Humanoids)Nava, G.; Romano F.; Nori F.; Pucci, D.; "Stability Analysis and Design of Momentum-based Controllers for Humanoid Robots" IEEE International Conference on Intelligent Robots and Systems (IROS). 2016This video shows the latest results on the whole-body control of humanoid robots achieved by the Dynamic Interaction Control Lab at the Italian Institute of Technology. The control of the robot is achieved by regulating the interaction forces between the robot and its surrounding environment. The force and torque exchanged between the robot's feet and the floor is regulated so that the robot keeps its balance even when strongly perturbed.In particular, the control architecture is composed of two nested control loops. The internal loop, which runs at 1 KHz, is in charge of stabilizing any desired joint torque. This task is achieved thanks to an off-line identification procedure providing us with a reliable model of friction and motor constants. The outer loop, which generates desired joint torques at 100 Hz, is a momentum based control algorithm with the formalism of free-floating systems subject to constraints (i.e. Differential Algebraic Equation frameworks). More precisely, the control objective for the outer loop is the stabilization of the robot’s linear and angular momentum and the associated zero dynamics. The latter objective can be used to stabilize a desired joint configuration. The stability of the control framework is shown to be in the sense of Lyapunov. The contact forces and torques at the contacts are regulated so as to break the contact only at desired configurations. Switching between several contacts is taken into account thanks to a finite-state-machine that dictates the constraints acting on the system. The control framework is implemented on the iCub humanoid robot.Related projects:http://codyco.eu/http://orb.iwr.uni-heidelberg.de/koroibot/Useful links:http://www.icub.orghttps://www.iit.it/research/lines/dynamic-interaction-controlhttps://www.iit.it
A Cartesian 6-DoF Gaze Controller for Humanoid Robots iCub HumanoidRobot 2016-05-04 | This video shows how we address the problem of controlling the 3D fixation point of a binocular, 6 Degrees-of-Freedom (DOF), anthropomorphic head.It is possible to define the fixation point as the virtual end-effector of the kinematic chain composed by the neck and the eyes. Consequently,the control of the fixation point can be achieved using techniques for inverse kinematics and trajectory generation normally adopted for controllingrobotic arms. Further, the redundancy of the task allows for the integration of different corollary behaviors in addition to the main control loop:vestibulo-ocular reflex (VOR), sacccadic behavior, and gaze stabilization (for a video on the gaze stabilization system, please refer toyoutu.be/NSGea-tCLZM)REFERENCE PAPER: Roncone A., Pattacini U., Metta G., Natale L. 2016, 'A Cartesian 6-DoF Gaze Controller for Humanoid Robots',Proceedings of Robotics: Science and Systems (RSS), Ann Arbor, MI, June 18-22 2016
Simulated iCub gets real: repeatability of simulation results on the real platform. iCub HumanoidRobot 2016-04-28 | The video shows the results of the iCub yoga++ demo on simulation (gazebo) and on the real robot. The simulation can be replicated on Linux and OsX.1. Instruction on how the install the entire software framework can be found here: github.com/robotology/codyco-superbuild2. Instructions on how to run the specific demonstration can be found here: github.com/robotology-playground/WBI-Toolbox-controllers/#wbi-toolbox-controllers3. The scientific paper describing the controller and the software architecture is openly available here: http://journal.frontiersin.org/article/10.3389/frobt.2015.00006/full
iCub performing YOGA++ demo iCub HumanoidRobot 2016-04-28 | The video shows the iCub yoga++ demo. This simulation can be replicated on Linux and OsX.1. Instruction on how the install the entire software framework can be found here: github.com/robotology/codyco-superbuild2. Instructions on how to run the specific demonstration can be found here: github.com/robotology-playground/WBI-Toolbox-controllers/#wbi-toolbox-controllers3. The scientific paper describing the controller and the software architecture is openly available here: http://journal.frontiersin.org/article/10.3389/frobt.2015.00006/full
iCub yoga++ in simulation iCub HumanoidRobot 2016-04-28 | The video shows the results of the iCub yoga++ demo on simulation (gazebo). This simulation can be replicated on Linux and OsX.1. Instruction on how the install the entire software framework can be found here: github.com/robotology/codyco-superbuild2. Instructions on how to run the specific demonstration can be found here: github.com/robotology-playground/WBI-Toolbox-controllers/#wbi-toolbox-controllers3. The scientific paper describing the controller and the software architecture is openly available here: http://journal.frontiersin.org/article/10.3389/frobt.2015.00006/full
iCub getting better in balancing on one foot iCub HumanoidRobot 2015-11-10 | This video shows some of the work going on at the Italian Institute of Technology aimed at improving the capacities of iCub when balancing on one foot.
Learning Peripersonal Space on the iCub iCub HumanoidRobot 2015-09-21 | In this video, the tactile system is used in order to build a representation of space immediately surrounding the body - peripersonal space. In particular, the iCub skin acts as a reinforcement for the visual system, with the goal of enhancing the perception of the surrounding world. By exploiting a temporal and spatial congruence between a purely visual event (e.g. an object approaching the robot’s body) and a purely tactile event (e.g. the same object eventually touching a skin part), a representation is learned that allows the robot to autonomously establish a margin of safety around its body through interaction with the environment - extending its cutaneous tactile space into the space surrounding it.We considered a scenario where external objects were approaching individual skin parts. A volume was chosen to demarcate a theoretical visual “receptive field” around every taxel. Learning then proceeded in a distributed, event-driven manner - every taxel stores and continuously updates a record of the count of positive (resulting in contact) and negative examples it has encountered.
Preliminary results on iCub balancing on a seesaw iCub HumanoidRobot 2015-07-19 | This video shows very preliminary results achieved in the whole-body control of iCub when standing on a seesaw. The knowledge of the robot and seesaw dynamics along with the measurement of the external perturbations allow him to keep the equilibrium even when interacting with humans.Well, experiments still do not work as well as simulations, but we are coming along nicely!The results have been achieved by the the researches funded by the European Project CoDyCo.
The iCub audio and visual attention system iCub HumanoidRobot 2015-07-04 | Saliency based sensor fusion of the audio and visual channels into a unique saliency map. Additional gaze strategy directs the iCub`s head towards stimuli off the field of view.Project Related:Codefror (https://www.codefror.eu/)Publications Related:"Saliency Based Sensor Fusion of Broadband Sound Localizer for Humanoids" the 2015 IEEE International Conference on Multisensor Fusion and Information Integration (MFI 2015), Sept 14-16, 2015, San Diego, CA
The iCub Project: a shared platform for research in artificial intelligence and robotics iCub HumanoidRobot 2015-06-23 | ...
