iFeel, wearable technology for human health monitoring Artificial and Mechanical Intelligence 2024-09-13 | #iFeel aims at improving people’s lives by developing #wearable #technologies that monitor the #biomechanical #risks of future workers and monitor key-performance indicators of future #rehabilitation patients.Research and development focus on two fields.REHABILITATION: This involves the process and analysis of Key Performance Indicators (KPIs) that are specifically relevant to mobility rehabilitation. PREVENTION: iFeel conducts real-time analysis and prediction of biomechanical risk indices with early haptic feedback for posture correction and risk reduction. Furthermore, iFeel also enables human-robot collaboration tasks and teleoperation of humanoid robot avatars. iFeel was first developed within the Horizon 2020 EU An.Dy. project and is now carried out in the ergoCub (www.ergocub.eu) project sponsored by the National Institute for Insurance against Accidents at Work (INAIL).Find out more on www.ifeeltech.eu
From CAD to URDF: Co-Design of a Jet-Powered Humanoid RobotIncluding CAD Geometry Artificial and Mechanical Intelligence 2024-10-12 | pre-print: arxiv.org/abs/2410.07963Abstract—Co-design optimization strategies usually rely onsimplified robot models extracted from CAD. While thesemodels are useful for optimizing geometrical and inertialparameters for robot control, they might overlook importantdetails essential for prototyping the optimized mechanicaldesign. For instance, they may not account for mechanicalstresses exerted on the optimized geometries and the complexityof assembly-level design. In this paper, we introduce a co-designframework aimed at improving both the control performanceand mechanical design of our robot. Specifically, we identifythe robot links that significantly influence control performance.The geometric characteristics of these links are parameterizedand optimized using a multi-objective evolutionary algorithmto achieve optimal control performance. Additionally, an automatedFinite Element Method (FEM) analysis is integratedinto the framework to filter solutions not satisfying the requiredstructural safety margin. We validate the framework by applyingit to enhance the mechanical design for flight performanceof the jet-powered humanoid robot iRonCub.
Automatic Gain Tuning for Humanoid Robots Walking Architectures Using Gradient-Free Opt. Techniques Artificial and Mechanical Intelligence 2024-09-27 | pre-print: arxiv.org/abs/2409.18649abstract: Developing sophisticated control architectures hasendowed robots, particularly humanoid robots, with numerouscapabilities. However, tuning these architectures remains achallenging and time-consuming task that requires expertintervention. In this work, we propose a methodology toautomatically tune the gains of all layers of a hierarchicalcontrol architecture for walking humanoids. We tested ourmethodology by employing different gradient-free optimizationmethods: Genetic Algorithm (GA), Covariance Matrix Adap-tation Evolution Strategy (CMA-ES), Evolution Strategy (ES),and Differential Evolution (DE). We validated the parameterfound both in simulation and on the real ergoCub humanoidrobot. Our results show that GA achieves the fastest convergence(10 × 10^3 function evaluations vs 25 × 10^3 needed by the otheralgorithms) and 100% success rate in completing the task bothin simulation and when transferred on the real robotic platform.These findings highlight the potential of our proposed methodto automate the tuning process, reducing the need for manualintervention.
Learning to Walk and Fly with Adversarial Motion Priors (IROS 2024) Artificial and Mechanical Intelligence 2024-09-09 | Robot multimodal locomotion encompasses the ability to transition between walking and #flying , representing a significant challenge in robotics. This work presents an approach that enables automatic smooth transitions between legged and aerial locomotion. Leveraging the concept of Adversarial Motion Priors, our method allows the robot to imitate motion datasets and accomplish the desired task without the need for complex reward functions. The robot learns walking patterns from human-like gaits and aerial locomotion patterns from motions obtained using trajectory optimization.Through this process, the robot adapts the locomotion scheme based on #environmental #feedback using reinforcement learning, with the spontaneous emergence of mode-switching behavior.The results highlight the potential for achieving #multimodal #locomotion in #aerial #humanoid #robotics through automatic control of walking and flying modes, paving the way for applications in diverse domains such as search and rescue, surveillance, and exploration missions.This research contributes to advancing the capabilities of aerial humanoid #robots in terms of versatile locomotion in various environments.The platform used is #iRonCub, based on the robot #iCub.Reference:Giuseppe L'Erario, Drew Hanover, Angel Romero, Yunlong Song, Gabriele Nava, Paolo Maria Viceconte, Daniele Pucci, Davide Scaramuzza"Learning to Walk and Fly with Adversarial Motion Priors"IEEE/RSJ International Conference on Intelligent Robots and Systems, Abu Dhabi, 2024Paper: arxiv.org/abs/2309.12784For more info about our research, check out our webpages:- Artificial and Mechanical Intelligence Lab: https://ami.iit.it/it/- Robotics and Perception Group: https://rpg.ifi.uzh.ch/- iRonCub Project: https://ami.iit.it/it/aerial-humanoid-roboticsAffiliations:Giuseppe L'Erario. Gabriele Nava, Paolo Maria Viceconte and Daniele Pucci are Artificial Mechanical Intelligence Lab, Istituto Italiano di Tecnologia, Genova, Italy.Drew Hanover, Angel Romero, Yunlong Song and Davide Scaramuzza are with the Robotics and Perception Group, Dep. of Informatics, University of Zurich, and Dep. of Neuroinformatics, University of Zurich and ETH Zurich, Switzerland.
iRonCub3 sneak peek: experimental area and preliminary validations of a jet-powered humanoid robot Artificial and Mechanical Intelligence 2024-08-12 | #controllersettings #ft #jetpacksSo far, we are the only research unit working on #flying #jetpowered #humanoidrobotics; for real. The video below then shows a sneak peek of the work we have been doing over the last two years to implement and test a new jet-powered humanoid robot, the #iRonCub3.The complexity of this research axis much differs from the classical challenges of humanoid robotics: #thermodynamics plays a pivotal role, since turbine emission gas is at about 800 degrees celsius and at almost the speed of sound; #aerodynamics of multi-body systems needs #neuralnetworks with #physicsinformed components to be evaluated on-line; #controllersettings need to combine high- and low-bandwidth actuators as joints and turbines; #planners are required to generate not only motor dynamics but also turbine trajectories; the #experimental validation is as serious as dangerous, so there is little room for improvisations. Right now, we are in the process of testing iRonCub3 in our new flight and control area: a big leap forward than what we did for the iRonCub2 (see IEEE Spectrumhttps://spectrum.ieee.org/flying-huma.... In particular, iRonCub3 surpasses its previous version in a number of features: the robot is based on iCub3 - see https://www.science.org/doi/10.1126/s... - so #tendons have been removed, #ft sensors integrated in the #jetpacks , new #electronics was designed, #control and #planners are of a new generation at higher frequency. Although challenging and difficult to dictate timelines, we are confident that this new robot and setup will be more effective during hovering. So, stay tuned.I am truly thankful to the researchers and engineers working on this research axis, and without whom the associated complexity would be unmanageable. First, Gabriele Nava, scrum master of the iRonCub team dealing with integration, management, and procedures, and then (alphabetical order): Fabio Bergonti, mechanics and control, Francesca Bruzzone, video and media, Davide Gorbani, on-line control and embedded systems, Giuseppe L'Erario, trajectory planning and estimation, Hosameldin A. O. Mohamed, thrust force estimation and FT modelling, Antonello Paolino, aerodynamics modelling and control, Shabarish Purushothaman Pillai, mechanics and integration, Saverio Taliani, mpc and validation, Vpunith Reddy, mechanics and co-design, Silvio Traversaro, software architectures and modelling, Nicholas Tremaroli, telemetry and estimation.This is possible thanks only to the entire Artificial and Mechanical Intelligence and the Istituto Italiano di Tecnologia, which keeps helping in making impossible ideas come true.
iRonCub3 sneak peek: experimental area and preliminary validations Artificial and Mechanical Intelligence 2024-07-25 | So far, we are the only research unit working on #flying #jetpowered #humanoidrobotics; for real. The video below then shows a sneek peek of the work we have been doing over the last two years to implement and test a new jet-powered humanoid robot, the #iRonCub3.The complexity of this research axis much differs from the classical challenges of humanoid robotics: #thermodynamics plays a pivotal role, since turbine emission gas is at about 800 degrees celsius and at almost the speed of sound; #aerodynamics of multi-body systems needs #neuralnetworks with #physicsinformed components to be evaluated on-line; #controllersettings need to combine high- and low-bandwidth actuators as joints and turbines; #planners are required to generate not only motor dynamics but also turbine trajectories; the hashtag#experimental validation is as serious as dangerous, so there is little room for improvisations. Right now, we are in the process of testing iRonCub3 in our new flight and control area: a big leap forward than what we did for the iRonCub2 (see IEEE Spectrumspectrum.ieee.org/flying-humanoid-robot). In particular, iRonCub3 surpasses its previous version in a number of features: the robot is based on iCub3 - see science.org/doi/10.1126/scirobotics.adh3834 - so #tendons have been removed, #ft sensors integrated in the #jetpacks , new #electronics was designed, #control and #planners are of a new generation at higher frequency. Although challenging and difficult to dictate timelines, we are confident that this new robot and setup will be more effective during hovering. So, stay tuned.I am truly thankful to the researchers and engineers working on this research axis, and without whom the associated complexity would be unmanageable. First, Gabriele Nava, scrum master of the iRonCub team dealing with integration, management, and procedures, and then (alphabetical order): Fabio Bergonti, mechanics and control, Francesca Bruzzone, video and media, Davide Gorbani, on-line control and embedded systems, Giuseppe L'Erario, trajectory planning and estimation, Hosameldin A. O. Mohamed, thrust force estimation and FT modelling, Antonello Paolino, aerodynamics modelling and control, Shabarish Purushothaman Pillai, mechanics and integration, Saverio Taliani, mpc and validation, Vpunith Reddy, mechanics and co-design, Silvio Traversaro, software architectures and modelling, Nicholas Tremaroli, telemetry and estimation.This is possible thanks only to the entire Artificial and Mechanical Intelligence and the Istituto Italiano di Tecnologia, which keeps helping in making impossible ideas come true.
XBG: End-to-end Imitation Learning for Autonomous Behaviour in Human-Robot Interaction Collaboration Artificial and Mechanical Intelligence 2024-06-25 | This paper presents XBG (eXteroceptive Behaviour Generation), a multimodal end-to-end Imitation Learning (IL) system for a whole-body autonomous humanoid robot used in real-world Human-Robot Interaction (HRI) scenarios. The main contribution of this paper is an architecture for learning HRI behaviors using a data-driven approach. Through teleoperation, a diverse dataset is collected, comprising demonstrations across multiple HRI scenarios, including handshaking, handwaving, payload reception, walking, and walking with a payload. After synchronizing, filtering, and transforming the data, different Deep Neural Networks (DNN) models are trained. The final system integrates different modalities comprising exteroceptive and proprioceptive sources of information to provide the robot with an understanding of its environment and its own actions. The robot takes sequence of images (RGB and depth) and joints state information during the interactions and then reacts accordingly, demonstrating learned behaviors. By fusing multimodal signals in time, we encode new autonomous capabilities into the robotic platform, allowing the understanding of context changes over time. The models are deployed on ergoCub, a real-world humanoid robot, and their performance is measured by calculating the success rate of the robot's behavior under the mentioned scenarios.
UKF-Based Sensor Fusion for Joint-Torque Sensorless Humanoid Robots Artificial and Mechanical Intelligence 2024-05-06 | arXiv: arxiv.org/abs/2402.18380Code link: github.com/ami-iit/paper_sorrentino_2024_icra_robot-dynamics-estimationAbstractThis paper proposes a novel sensor fusion based on Unscented Kalman Filtering for the online estimation of joint-torques of humanoid robots without joint-torque sensors. At the feature level, the proposed approach considers multimodal measurements (e.g. currents, accelerations, etc.) and non-directly measurable effects, such as external contacts, thus leading to joint torques readily usable in control architectures for human-robot interaction. The proposed sensor fusion can also integrate distributed, non-collocated force/torque sensors, thus being a flexible framework with respect to the underlying robot sensor suit. To validate the approach, we show how the proposed sensor fusion can be integrated into a twolevel torque control architecture aiming at task-space torquecontrol. The performances of the proposed approach are shown through extensive tests on the new humanoid robot ergoCub, currently being developed at Istituto Italiano di Tecnologia. We also compare our strategy with the existing state-of-theart approach based on the recursive Newton-Euler algorithm. Results demonstrate that our method achieves low root mean square errors in torque tracking, ranging from 0.05 Nm to 2.5 Nm, even in the presence of external contacts.
Co-Design Optimisation of Morphing Topology and Control of Winged Drones [7min] Artificial and Mechanical Intelligence 2024-05-06 | Co-Design Optimisation of Morphing Topology and Control of Winged Dronesshort video: youtu.be/uWYuQ8gT404arXiv link: arxiv.org/abs/2309.13948Code link: github.com/ami-iit/paper_bergonti_2024_icra_codesign-morphing-dronesLab link: https://ami.iit.it/Abstract: The design and control of winged aircraft and drones is an iterative process aimed at identifying a compromise of mission-specific costs and constraints. When agility is required, shape-shifting (morphing) drones represent an efficient solution. However, morphing drones require the addition of actuated joints that increase the topology and control coupling, making the design process more complex. We propose a co-design optimisation method that assists the engineers by proposing a morphing drone’s conceptual design that includes topology, actuation, morphing strategy, and controller parameters. The method consists of applying multi-objective constraint-based optimisation to a multi-body winged drone with trajectory optimisation to solve the motion intelligence problem under diverse flight mission requirements, such as energy consumption and mission completion time. We show that co-designed morphing drones outperform fixed-winged drones in terms of energy efficiency and mission time, suggesting that the proposed co-design method could be a useful addition to the aircraft engineering toolbox.
Real-time Lower Leg Muscle Forces Estimation using a Hill-type Model and Whole-body Wearable Sensors Artificial and Mechanical Intelligence 2024-03-29 | Abstract: The estimation of the forces produced by muscles is an aspect of great interest in several fields, e.g., rehabilitation, human-machine interface and sports, where it is pivotal to estimate the muscle forces produced by the human biomechanics control system. Since muscles are the engines of human motion, the scientific community is actively involved in studying methods to model muscles in order to simulate human motions. In our preliminary study, we built a computational model for the contraction dynamics estimation of three muscles of the lower leg (tibialis anterior, gastrocnemius lateralis, gastrocnemius medialis). Acquisition data have been gathered by using the iFeel technology https://ifeeltech.eu (IMU nodes and sensorized portable shoes) developed by the Artificial and Mechanical Intelligence (AMI) of the Italian Institute of Technology and a off-the-shelf system of surface electromyography (EMGs). The architecture has been implemented on the middleware YARP for both the in-lab sensors and the EMGs tool.
Co-Design Optimisation of Morphing Topology and Control of Winged Drones Artificial and Mechanical Intelligence 2024-03-05 | Co-Design Optimisation of Morphing Topology and Control of Winged Dronesfull video: youtu.be/X_iAKJx8ml0arXiv link: arxiv.org/abs/2309.13948Code link: github.com/ami-iit/paper_bergonti_2024_icra_codesign-morphing-dronesLab link: https://ami.iit.it/Abstract: The design and control of winged aircraft and drones is an iterative process aimed at identifying a compromise of mission-specific costs and constraints. When agility is required, shape-shifting (morphing) drones represent an efficient solution. However, morphing drones require the addition of actuated joints that increase the topology and control coupling, making the design process more complex. We propose a co-design optimisation method that assists the engineers by proposing a morphing drone’s conceptual design that includes topology, actuation, morphing strategy, and controller parameters. The method consists of applying multi-objective constraint-based optimisation to a multi-body winged drone with trajectory optimisation to solve the motion intelligence problem under diverse flight mission requirements, such as energy consumption and mission completion time. We show that co-designed morphing drones outperform fixed-winged drones in terms of energy efficiency and mission time, suggesting that the proposed co-design method could be a useful addition to the aircraft engineering toolbox.
Invited talk: Safe robot control combining learning and model predictive control by Andrea Del Prete Artificial and Mechanical Intelligence 2024-02-22 | Abstract In recent years, advanced model-based and data-driven control methods are unlocking the potential of complex robotics systems, and we can expect this trend to continue at an exponential rate in the near future. However, ensuring safety with these advanced control methods remains a challenge. A well-known tool to make controllers (either Model Predictive Controllers or Reinforcement Learning policies) safe, is the so-called control-invariant set (a.k.a. safe set). Unfortunately, for nonlinear systems, such a set cannot be exactly computed in general. Numerical algorithms exist for computing approximate control-invariant sets, but classic theoretic control methods break down if the set is not exact. In this presentation I will discuss our recent efforts to address this issue. First, we have designed a novel algorithm for computing a numerical approximation of the largest control invariant set for a robot manipulator. This method is more efficient than other state-of-the-art approaches because it exploits some properties of the dynamics of fully-actuated multi-body systems. Second, I will present a novel Model Predictive Control scheme that can guarantee recursive feasibility and/or safety under weaker assumptions than classic methods. In particular, recursive feasibility is guaranteed by making the safe-set constraint move backward over the horizon, and assuming that such set satisfies a condition that is weaker than control invariance. Safety is instead guaranteed under an even weaker assumption on the safe set, triggering a safe task-abortion strategy whenever a risk of constraint violation is detected. We evaluated our approach on a simulated robot manipulator, empirically demonstrating that it leads to less constraint violations than state-of-the-art approaches, while retaining good performance in terms of tracking cost, number of completed tasks, and computation times.Bio Since 2022 Andrea has been an associate professor in the Industrial Engineering Department of the University of Trento (Italy). From 2019 to 2021 he had been a tenure-track assistant professor (RTD-B) in the same department. In 2018 he had been a research scientist at the Max-Planck Institute for Intelligent Systems (Tübingen, Germany). From 2014 to 2017 he had been an associated researcher at LAAS/CNRS, in Toulouse (France), where he has been working with the humanoid robot HRP-2. Before going to LAAS he had spent four years (3 of PhD + 1 of post-doc) at the Italian Institute of Technology (IIT, Genova, Italy), working on the iCub humanoid robot. His research interests lie at the intersection between control, optimization and machine learning.
Invited Talk - From basic studies to industrial applications with 2D materials, Camilla Coletti, IIT Artificial and Mechanical Intelligence 2024-02-16 | Title "Moving from fundamental studies to industrial applications with 2D materials"Biography: Camilla Coletti is a tenured Senior Scientist of the Istituto Italiano di Tecnologia (IIT) and principal investigator of the research line 2D Materials Engineering: She is the coordinator of the Center for Nanotechnology Innovation (CNI@NEST) of Pisa and of the Graphene Labs. She has been hired by IIT in 2011 after being an Alexander von Humboldt postdoctoral fellow at the Max Planck Institute for Solid State Research of Stuttgart (Germany). She received her PhD degree from the University of South Florida in 2007 and her MS degree from the University of Perugia in 2004 (with honors, both in Electrical Engineering). She is expert in the synthesis of highly-crystalline 2D materials via chemical vapour deposition (CVD) and in the investigation of their electronic, chemical and structural properties. Her research is focused on: (i) synthesis and integration of scalable 2D materials for optoelectronics, photonics and biomedicine; (ii) engineering the interface and properties of 2D heterostructures. In her work she applies her background of surface scientist to impact science and technology of 2D materials. Over the years, within IIT, she has directly supervised over 50 students and postdocs, with a great attention in creating an international and gender fair environment. She has received funding for more than 6.5 M€ from competitive grants and industrial contracts. In the last years she has been an invited speaker at prestigious conferences in the field (e.g., Graphene Week, Graphene 20xx, MRS Spring/Fall), she organized symposia and workshops (e.g., Japan-EU Flagship Workshops, MRS Symposia), and she was a lecturer at Summer/Winter Schools (e.g., Graphene Winter School, Nano-Network, Cambridge Advanced Technology Lectures). Since 2020 she is Faculty Board Member for the PhD in Nanoscience at the Scuola Normale Superiore (SNS) of Pisa and throughout the years she has been a lecturer for courses at the Master’s and PhD level (University of Pisa, University of Genova, SNS). Overall, she is author of more than 170 peer-reviewed publications, authored 4 book chapters, edited 1 book, filed several international patents (holds 3) and delivered more than 60 invited talks at international conferences.Abstract: Graphene displays a plethora of enticing properties such as exceptional electron and thermal conductivity, mechanical strength and flexibility, impermeability, and transparency, which make it an enticing candidate for a number of applications. Yet, to make graphene and other two-dimensional (2D) materials viable at high technology readiness levels, requirements such as high-quality, scalability and contamination control have to be satisfied. In this talk, I will discuss scalable growth of high-quality graphene and transition metal dichalcogenides via chemical vapor deposition (CVD) [1-3] and discuss how these scalable 2D materials can be adopted in industrial applications, from integrated optical photonics [4], to Hall sensors, biomedicine [5] and power distribution [6]. Also, the potential of 2D materials for quantum technology will be discussed.References1.M.A. Giambra, V. Mišeikis, S. Pezzini, S. Marconi, A. Montanaro, F. Fabbri, V. Sorianello, A.C. Ferrari, C. Coletti, M. Romagnoli, ACS nano 15 (2), 2021.2.N Mishra, S Forti, F Fabbri, L Martini, C McAleese, B Conran, PR Whelan, A Shivayogimath, BS Jessen, L Buß, J Falta, I Aliaj, S Roddaro, JI Flege, P Bøggild, KBK Teo, C Coletti, Small 15 (50), 2019.3.S. Pace, L. Martini, D. Convertino, D.-H. Keum, S. Forti, S. Pezzini, F. Fabbri, V. Mišeikis, C. Coletti, ACS nano 15 (3), 2021.4.A. Montanaro, G. Piccinini, V. Mišeikis, V. Sorianello, M.A. Giambra, S. Soresi, L. Giorgi, A. D’Errico, K. Watanabe, T. Taniguchi, S. Pezzini, C. Coletti, M. Romagnoli, just accepted, Nature Communications 2023 assets.researchsquare.com/files/rs-1835036/v1/b71c7e98-a56d-4193-9e58-0b1fd485cfc1.pdf?c=16596487875.Domenica Convertino, Filippo Fabbri, Neeraj Mishra, Marco Mainardi, Valentina Cappello, Giovanna Testa, Simona Capsoni, Lorenzo Albertazzi, Stefano Luin, Laura Marchetti, Camilla Coletti, Nano Letters 20 (5), 3633-3641 20206.N. Mishra, Y. Vlamidis, L. Martini, A. Lanza, A. Jouvray, M. La Sala, M. Gemmi, V. Mišeikis, M. Perry, K.B.K. Teo, S. Forti, C. Coletti, ACS Appl. Eng. Mater. 1, 7, 1937–1945, 2023
iCub3 Avatar System: Enabling Remote Fully-Immersive Embodiment of Humanoid Robots Artificial and Mechanical Intelligence 2024-01-25 | Video associated to the paper "iCub3 Avatar System: Enabling Remote Fully-Immersive Embodiment of Humanoid Robots" published in Science Robotics. Link to publication: science.org/doi/10.1126/scirobotics.adh3834 Authors: Stefano Dafarra, Ugo Pattacini, Giulio Romualdi, Lorenzo Rapetti, Riccardo Grieco, Kourosh Darvish, Gianluca Milani, Enrico Valli, Ines Sorrentino, Paolo Maria Viceconte, Alessandro Scalzo, Silvio Traversaro, Carlotta Sartore, Mohamed Elobaid, Nuno Guedelha, Connor Herron, Alexander Leonessa, Francesco Draicchio, Giorgio Metta, Marco Maggiali, Daniele Pucci We present an avatar system designed to facilitate the embodiment of humanoid robots by human operators, validated through iCub3, a humanoid developed at the Istituto Italiano di Tecnologia (IIT). More precisely, the contribution of the paper is twofold: first, we present the humanoid iCub3 as a robotic avatar which integrates the latest significant improvements after about fifteen years of development of the iCub series; second, we present a versatile avatar system enabling humans to embody humanoid robots encompassing aspects such as locomotion, manipulation, voice, and face expressions with comprehensive sensory feedback including visual, auditory, haptic, weight, and touch modalities. We validate the system by implementing several avatar architecture instances, each tailored to specific requirements. First, we evaluated the optimized architecture for verbal, non-verbal, and physical interactions with a remote recipient. This testing involved the operator in Genoa and the avatar in the Biennale di Venezia, Venice – about 290 Km away – thus allowing the operator to visit remotely the Italian art exhibition. Second, we evaluated the optimised architecture for recipient physical collaboration and public engagement on-stage, live, at the We Make Future show, a prominent world digital innovation festival. In this instance, the operator was situated in Genoa while the avatar operates in Rimini – about 300 Km away – interacting with a recipient who entrusted the avatar a payload to carry on stage before an audience of approximately 2000 spectators. Third, we present the architecture implemented by the iCub Team for the ANA Avatar XPrize competition.Check our social media:linkedin.com/company/ami-artificial-and-mechanical-intelligencetwitter.com/ami_iitinstagram.com/artificialmechintelligencefacebook.com/ArtificialMechIntelligence
Invited Talk - Computational Optimization in Legged Robot Co Design by Gabriele Fadini, LAAS-CNRS Artificial and Mechanical Intelligence 2024-01-16 | Abstract: Co-design is a paradigm to determine the best design of a robot for a given task. The focus of my research at LAAS-CNRS has been on high-performance and efficient-legged robotic systems' optimization. A framework, allowing a simultaneous optimization of hardware and task, has been proposed and used for different application scenarios and robot topologies. This method is based on splitting the problem in a bi-level optimization scheme, which features a genetic algorithm in the outer loop (considering hardware and task variables) and trajectory optimization in the inner loop (for the optimal trajectories). With this approach, simultaneous hardware and control optimization is achieved thanks to a complete robot model, accounting for the actuator losses and dynamics, and the structural scaling of the platform. This formulation is rather general and can be adapted to account for tasks, robots, and control policies. Biography: Gabriele Fadini was a Ph.D. student at LAAS-CNRS, under the supervision of Philippe Souères and Thomas Flayols. His research activity revolved around the co-design optimization of legged robots and optimal control. During the summer of 2022, he has been visiting researcher at DFKI - Bremen in the Underactuated Robotics Lab led by Dr. Shivesh Kumar for a project involving hardware co-optimization. Currently he is a Post-doctoral researcher at ETHZ, Computational Robotics Lab (CRL).
Codesign of Humanoid Robots for Ergonomy Collaboration with M. Humans via Gen. Algo. and Nonlin. Opt Artificial and Mechanical Intelligence 2024-01-14 | Paper: ieeexplore.ieee.org/document/10375237Preprint: arxiv.org/abs/2312.07459Abstract: Ergonomics is a key factor to consider when designing control architectures for effective physical collaborations between humans and humanoid robots. In contrast, ergonomic indexes are often overlooked in the robot design phase, which leads to suboptimal performance in physical human-robot interaction tasks. This paper proposes a novel methodology for optimizing the design of humanoid robots with respect to ergonomic indicators associated with the interaction of multiple agents. Our approach leverages a dynamic and kinematic parameterization of the robot link and motor specifications to seek for optimal robot designs using a bilevel optimization approach. Specifically, a genetic algorithm first generates robot designs by selecting the link and motor characteristics. Then, we use nonlinear optimization to evaluate interaction ergonomy indexes during collaborative payload lifting with different humans and weights. To assess the effectiveness of our approach, we compare the optimal design obtained using bilevel optimization against the design obtained using nonlinear optimization. Our results show that the proposed approach significantly improves ergonomics in terms of energy expenditure calculated in two reference scenarios involving static and dynamic robot motions. We plan to apply our methodology to drive the design of the ergoCub2 robot, a humanoid intended for optimal physical collaboration with humans in diverse environments.
Online Action Recognition for Human Risk Prediction with Anticipated Haptic Alert via Wearables Artificial and Mechanical Intelligence 2024-01-10 | Preprint:Paper: ieeexplore.ieee.org/document/10375149Abstract: This paper proposes a framework that combines online human state estimation, action recognition and motion prediction to enable early assessment and prevention of worker biomechanical risk during lifting tasks. The framework leverages the NIOSH index to perform online risk assessment, thus fitting realtime applications. In particular, the human state is retrieved via inverse kinematics/dynamics algorithms from wearable sensor data. Human action recognition and motion prediction are achieved by implementing an LSTM-based Guided Mixture of Experts architecture, which is trained offline and inferred online. With the recognized actions, a single lifting activity is divided into a series of continuous movements and the Revised NIOSH Lifting Equation can be applied for risk assessment. Moreover, the predicted motions enable anticipation of future risks. A haptic actuator, embedded in the wearable system, can alert the subject of potential risk, acting as an active prevention device. The performance of the proposed framework is validated by executing real lifting tasks, while the subject is equipped with the iFeel wearable system. The source code for this paper is available at github.com/ami-iit/paper-guo_2023_humanoids_lifting_risk.prediction.
Revolutionizing Work Safety: The ergoCub Project Artificial and Mechanical Intelligence 2024-01-08 | ergoCub is a project that develops technologies to catalyze the necessary digital transformation for reducing the number of musculoskeletal diseases related to biomechanical risk in future workers. This objective is pursued by developing wearable technologies, humanoid robots, and artificial intelligence while monitoring the acceptability of these technologies.The project is a collaboration between INAIL and the Artificial and Mechanical Intelligence Lab from IIT.Check our website https://ergocub.eu/Check our social media:linkedin.com/company/ami-artificial-and-mechanical-intelligencetwitter.com/ami_iitinstagram.com/artificialmechintelligencefacebook.com/ArtificialMechIntelligence
Nonlinear In-situ Calibration of Strain-Gauge Force/Torque Sensors for Humanoid Robots Artificial and Mechanical Intelligence 2024-01-06 | ieeexplore.ieee.org/document/10375227Abstract: High force/torque (F/T) sensor calibration accuracy is crucial to achieving successful force estimation/control tasks with humanoid robots. State-of-the-art affine calibration models do not always approximate correctly the physical phenomenon of the sensor/transducer, resulting in inaccurate F/T measurements for specific applications such as thrust estimation of a jet-powered humanoid robot. This paper proposes and validates nonlinear polynomial models for F/T calibration, increasing the number of model coefficients to minimize the estimation residuals. The analysis of several models, based on the data collected from experiments with the iCub3 robot, shows a significant improvement in minimizing the force/torque estimation error when using higher-degree polynomials. In particular, when using a 4th-degree polynomial model, the Root Mean Square error (RMSE) decreased to 2.28N from the 4.58N obtained with an affine model, and the absolute error in the forces remained under 6N while it was reaching up to 16N with the affine model.
Humanoid robot ergoCub walks and interacts with people at the 20 year anniversary of IIT Artificial and Mechanical Intelligence 2023-12-04 | For the first time in Italy 🇮🇹, ergoCub , our humanoid robot, walks during the demo for the 20 year anniversary of Istituto Italiano di Tecnologia. The robot was untethered, in front of about 400 spectators, and controlled remotely with the operator at about one kilometer. The humanoid robot is developed in the context of the ergoCub project (www.ergocub.eu) sponsored by INAIL, Istituto Nazionale Assicurazione contro gli Infortuni sul Lavoro, with the contribution of the Department of Medicine, Epidemiology, Occupational and Environmental Hygiene (DiMEILA).
Artificial and Mechanical Intelligence 14th Research increment Artificial and Mechanical Intelligence 2023-08-10 | Every quarter, we come #together to celebrate our #accomplishments and set our sights on #future #objectives. We believe staying in sync with the team is vital for driving our research forward!00:00 Daniele Pucci - Team update09:45 Giuseppe L'Erario - Learning to walk and fly with DeepRL24:57 S-Dafarra Alpha team: planned vs achieved objectives and statistics35:40 Carlotta Sartore - Gradient-Free Method for Hardware Optimization58:29 Giulio Romualdi - Delta team: planned vs achieved objectives and statistics1:11:36 Paolo Viceconte - A bird's-eye view on Transformers1:34:50 Mohamed Elobaid - Tau team: planned vs achieved objectives and statistics1:41:43 Ehsan Ranjbari - Hand Motion Retargeting for Teleoperated Grasping Tasks1:57:05 Lorenzo Rapetti - ergoCub team: planned vs achieved objectives and statistics2:14:42 Evelyn D'Elia - Modular Trajectory Generation for Humanoid Locomotion using Supervised Learning2:29:20 Enrico Valli - iFeel team: planned vs achieved objectives and statistics2:44:15 Gianluca Milani - A lightweight, capacitive-based iFeel Shoe#Research #Team #ami #iit #ergoCub #iRonCub #iCub #artificialintelligence #mechanicalengineering #avatar #robot #humanoid #italy #update
Learning-based methods for planning and control of humanoid robots, Paolo Maria Viceconte PhDdefense Artificial and Mechanical Intelligence 2023-06-09 | Link: https://iris.uniroma1.it/retrieve/3791429a-5bcf-44b8-9d23-c56e0f68878e/Tesi_dottorato_Viceconte.pdfThis thesis investigates the application of learning-based techniques to the planning and control of humanoid locomotion. In particular, both deep reinforcement learning and deep supervised learning are considered to address humanoid locomotion tasks in a crescendo of complexity.
Failure Detection and Fault Tolerant Control of a Jet-Powered Flying Humanoid Robot Artificial and Mechanical Intelligence 2023-05-22 | Pre-Print: arxiv.org/pdf/2305.16075v1.pdfCode: github.com/ami-iit/paper_nava_2023_icra_fault-control-ironcubAbstract: Failure detection and fault tolerant control are fundamental safety features of any aerial vehicle. With the emergence of complex, multi-body flying systems such as jet-powered humanoid robots, it becomes of crucial importance to design fault detection and control strategies for these systems, too. In this paper we propose a fault detection and control framework for the flying humanoid robot iRonCub in case of loss of one turbine. The framework is composed of a failure detector based on turbines rotational speed, a momentum-based flight control for fault response, and an offline reference generator that produces far-from-singularities configurations and accounts for self and jet exhausts collision avoidance. Simulation results with Gazebo and MATLAB prove the effectiveness of the proposed control strategy.
Online Non-linear Centroidal MPC for Humanoid Robots Payload Carrying with Force Parametrization Artificial and Mechanical Intelligence 2023-05-19 | Pre-print: arxiv.org/abs/2305.10917Code: github.com/ami-iit/paper_elobaid_2023_icra_walking_with_payloadsAbstract: In this paper we consider the problem of allowing a humanoid robot that is subject to a persistent disturbance, in the form of a payload-carrying task, to follow given planned footsteps. To solve this problem, we combine an online nonlinear centroidal Model Predictive Controller - MPC with a contact stable force parametrization. The cost function of the MPC is augmented with terms handling the disturbance and regularizing the parameter. The performance of the resulting controller is validated both in simulations and on the humanoid robot iCub. Finally, the effect of using the parametrization on the computational time of the controller is briefly studied.
A Control Approach for Human Robot Ergonomic Payload Lifting Artificial and Mechanical Intelligence 2023-05-16 | Pre-print: arxiv.org/abs/2305.08499 Code: github.com/ami-iit/paper_rapetti_2023_icra_ergonomic_payload_liftingAbstract: Collaborative robots can relief human operators from excessive efforts during payload lifting activities. Modelling the human partner allows the design of safe and efficient collaborative strategies. In this paper, we present a control approach for human-robot collaboration based on human monitoring through whole-body wearable sensors, and interaction modelling through coupled rigid-body dynamics. Moreover, a trajectory advancement strategy is proposed, allowing for online adaptation of the robot trajectory depending on the human motion. The resulting framework allows us to perform payload lifting tasks, taking into account the ergonomic requirements of the agents. Validation has been performed in an experimental scenario using the iCub3 humanoid robot and a human subject sensorized with the iFeel wearable system.
Torque Control with Joints Position and Velocity Limits Avoidance Artificial and Mechanical Intelligence 2023-05-14 | Preprint of ICRA 2023: arxiv.org/pdf/2303.17252.pdfCode: github.com/ami-iit/paper_pasandi_2023_icra-joint-limit-avoidanceAbstract: This paper presents a control architecture for tracking the desired time-varying trajectory while ensuring the joints position and velocity limits for fully actuated manipulators. The presented architecture stems from the parametrization of the feasible joints position and velocity space throughintroduced states. The proposed parametrization transforms the control problem with constrained states to an unconstrained one byreplacing the joints position and velocity terms with theintroduced states. With the help of Lyapunov-based arguments, we prove that the proposed control architecture ensures the stability and convergence of the desired joint trajectory along with the joints position and velocity limits avoidance. The simulations on a simple two-degree-of-freedom manipulator and the humanoid robot iCub show the effectiveness of the proposed architecture.
Invited Talk - Active physical human robot interaction: what and how by Yue Hu, A.I.R. Lab Artificial and Mechanical Intelligence 2023-04-12 | Abstract: Two research directions have been trying to breach the barrier between humans and robots: physically (pHRI), and socially (sHRI), which have been evolving without many intersections. But for robots to really coexist and collaborate with humans, it is necessary to take into account both physical and social interactions to achieve active interaction. We envision active physical human-robot interaction (active pHRI) to be a type of interaction during the robot is no more passively adapting, but becomes an active collaborative partner by understanding not only its tasks, but also the human users' physical and mental state. In this talk, I will illustrate the experiments performed to target fundamental steps towards achieving active pHRI: understanding the relationships between the humans perceptions and their measurable data when a robot takes direct active physical actions on the human, with insights on outcomes, future applications, and developments.Biography:Yue Hu is Assistant Professor in the Department of Mechanical & Mechatronics Engineering at the University of Waterloo. Head of the Active & Interactive Robotics Laboratory (A.I.R. Lab).Her main research interests are social-physical human robot-interaction, optimization and optimal control for robots, collaborative robots, humanoid robots, and human motion analysis. #invited #talk #ami #iit #airlab #robots #humanoid #walking #interaction #phri
iCub3 teleoperated at Ministry of Health, Spine 4.0 event Artificial and Mechanical Intelligence 2023-04-11 | Our ergoCub team, with iCub3, participated in this event that took place in April 2022. The meeting promoted by the Italian Ministry of Health "Spine 4.0: innovation for the prevention and treatment of spinal pathologies" were introduced by the Minister of Health, Roberto Speranza, the Director General of Communication and European and International Relations of the Ministry of Health, Sergio Iavicoli, the General Director of INAIL, Andrea Tardiola, the President of SIOT, Paolo Tranquilli Leali and Giovanna Spatari, President SIML.Check the link for watch the full event:youtube.com/watch?v=6zip7D13MOE#iit #icub #artificialintelligence #mechanical #spine #ministry #health #italy #spinal #pathology #avatar #robotics #robot #humanoid #ergoCub #inail #prevention
Simultaneous Action Recognition & Human Whole-Body Motion & Dynamics Prediction from Wearable Sensor Artificial and Mechanical Intelligence 2023-04-06 | Abstract:This paper presents a novel approach to solve simultaneously the problems of human activity recognition and whole-body motion and dynamics prediction for real-time applications. Starting from the dynamics of human motion and motor system theory, the notion of mixture of experts from deep learning has been extended to address this problem. In the proposed approach, experts are modelled as a sequence-to-sequence recurrent neural networks (RNN) architecture. Experiments show the results of 66-DoF real-world human motion prediction and action recognition during different tasks like walking and rotating.Authors: Kourosh Darvish; Serena Ivaldi; Daniele PucciLinks:Paper: ieeexplore.ieee.org/document/10000122Preprint: arxiv.org/pdf/2303.07655v1.pdfGitHub repository: github.com/ami-iit/paper_darvish_2022_humanoids_action-kindyn-predicitionDataset: zenodo.org/record/7731386#.ZC6LFHZBwYq
MADE IN INAIL - Rome: the President of the Italian Republic shook the hand with our iCub3 Artificial and Mechanical Intelligence 2023-04-04 | The President of the Italian Republic Sergio Mattarella was in Rome and shook the hand with our #iCub3, teleoperated from Genoa more than 400 Kms away.It was an amazing experience and gave us the chance to show him our #iFeel wearable technology and a sneak peek of our #ergoCub joint project with #Inail.Thanks to the Made in Inail event for this great opportunity!Made in Inail 2022#technology #opportunity #event #experience #ami #artificialintelligence #mechanical #icub #avatar #teleoperation #robotics #robot #iit #research #ifeel #italy #project #italy #ergoCub
2023-03-24 Feedback linearization via digital control Artificial and Mechanical Intelligence 2023-03-24 | An insight into the feedback linearisation techniques via digital control techniques by scholar.google.com/citations?user=hmvmK8gAAAAJ&hl=it a post-doc at the https://ami.iit.it/
iRonCub preparation Artificial and Mechanical Intelligence 2023-03-20 | Setting up an experiment it's not a trivial matter, there are many things to check twice or more, and with a jet powered humanoid robot is even harder! Here you can see a glimpse of our #iRonCub team, during the usual set up before the tests!#ami #aerial #humanoid #robot #iCub #jet #powered #test #demonstration #artificialintelligence #mechanical #technology #robotics #engineer #team #iit
Whole-Body Trajectory Optimization for Robot Multimodal Locomotion Artificial and Mechanical Intelligence 2023-02-19 | Preprint: arxiv.org/pdf/2211.12849.pdfCode: github.com/ami-iit/paper_lerario_2022_humanoids_planning-multimodal-locomotionAbstract: The general problem of planning feasible trajectories for multimodal robots is still an open challenge. Thispaper presents a whole-body trajectory optimisation approachthat addresses this challenge by combining methods and toolsdeveloped for aerial and legged robots. First, robot modelsthat enable the presented whole-body trajectory optimisationframework are presented. The key model is the so-calledrobot centroidal momentum, the dynamics of which is directlyrelated to the models of the robot actuation for aerial andterrestrial locomotion. Then, the paper presents how thesemodels can be employed in an optimal control problem togenerate either terrestrial or aerial locomotion trajectorieswith a unified approach. The optimisation problem considersrobot kinematics, momentum, thrust forces and their bounds.The overall approach is validated using the multimodal robotiRonCub, a flying humanoid robot that expresses a degreeof terrestrial and aerial locomotion. To solve the associatedoptimal trajectory generation problem, we employ ADAM, acustom-made open-source library that implements a collectionof algorithms for calculating rigid-body dynamics using CasADi.#iit #ami #robotics #aerial #robot #artificialintelligence #iRonCub #iCub #trajectory #optimization #multimodal #locomotion
Optimization of Humanoid Robot Designs for Human-Robot Ergonomic Payload Lifting Artificial and Mechanical Intelligence 2023-02-13 | pre-print: arxiv.org/abs/2211.13503code: github.com/ami-iit/paper_sartore_2022_humanoids_ergonomic_designabstract: When a human and a humanoid robot collaborate physically, ergonomics is a key factor to consider. Assuming a given humanoid robot, several control architectures exist nowadays to address ergonomic physical human-robot collaboration. This paper takes one step further by considering robot hardware parameters as optimization variables in the problem of collaborative payload lifting. The variables that parametrize robot's kinematics and dynamics ensure their physical consistency, and the human model is considered in the optimization problem. By leveraging the proposed modelling framework, the ergonomy of the interaction is maximized, here given by the agents' energy expenditure. Robot kinematic, dynamics, hardware constraints and human geometries are considered when solving the associated optimization problem. The proposed methodology is used to identify optimum hardware parameters for the design of the ergoCub robot, a humanoid possessing a degree of embodied intelligence for ergonomic interaction with humans. For the optimization problem, the starting point is the iCub humanoid robot. The obtained robot design reaches loads at heights in the range of 0.8-1.5 m with respect to the iCub robot whose range is limited to 0.8-1.2 m. The robot energy expenditure is decreased by about 33%, meanwhile, the human ergonomy is preserved, leading overall to an improved interaction.
Invited talk: An overview of Research and Robotics at Google DeepMind by Francesco Nori, DeepMind Artificial and Mechanical Intelligence 2023-02-01 | AbstractDeepMind is working on some of the world’s most complex and interesting research challenges, with the ultimate goal of solving artificial general intelligence (AGI). We ultimately want to develop an AGI capable of dealing with a variety of environments. A truly general AGI needs to be able to act on the real world and to learn tasks on real robots. Robotics at DeepMind aims at endowing robots with the ability to learn how to perform complex manipulation and locomotion tasks. This talk will give an introduction to DeepMind with specific focus on robotics, control or reinforcement learning.Bio Francesco was born in Padova in 1976. He received his D.Eng. degree (highest honors) from the University of Padova (Italy) in 2002. During the year 2002 he was a member of the UCLA Vision Lab as a visiting student under the supervision of Prof. Stefano Soatto, University of California Los Angeles. During this collaboration period he started a research activity in the field of computational vision and human motion tracking. In 2003 Francesco Nori started his Ph.D. under the supervision of Prof. Ruggero Frezza at the University of Padova, Italy. During this period the main topic of his research activity was modular control with special attention on biologically inspired control structures. Francesco Nori received his Ph.D. in Control and Dynamical Systems from the University of Padova (Italy) in 2005. In the year 2006 he moved to the University of Genova and started his PostDoc at the laboratory for integrated advanced robotics (LiraLab), beginning a fruitful collaboration with Prof. Giorgio Metta and Prof. Giulio Sandini. In 2007 Francesco Nori moved to the Italian Institute of technology where in 2015 he was appointed Tenure Track Researcher of the Dynamic and Interaction Control research line. His research interests are currently focused on whole-body motion control exploiting multiple (possibly compliant) contacts. With Giorgio Metta and Lorenzo Natale he is one of the key researchers involved in the iCub development, with specific focus on control and whole-body force regulation exploiting tactile information. Francesco is currently coordinating the H2020-EU project An.Dy (id. 731540); in the past he has been involved in two FP7-EU projects: CoDyCo as coordinator and Koroibot as principal investigator. In 2017 Francesco joined Deepmind where he is collaborating with Raia Hadsell, Nando de Freitas, Martin Riedmiller and Dan Belov. His current interests seamlessly span robotics and artificial intelligence, with applications in both manipulation and locomotion.#robotics #google #deepmind #ami #iit #artificialintelligence #mechanical #robot #technology #research #agi
Happy 2023 from Artificial and Mechanical Intelligence Artificial and Mechanical Intelligence 2022-12-29 | Looking back on this year we are proud of what we have achieved and would like to share some of the best moments with you. We are thrilled to see what 2023 will bring us! We wish you a happy new year!#iit #ami #artificialintelligence #mechanical #holiday #happynewyear #icub #robotics #technology #achivements #challenges
Invited Talk: Motion control bits for homemade robots by Prof. Stéphane Caron Artificial and Mechanical Intelligence 2022-12-13 | We have learned a lot from big expensive robots, but their weight and price are likely damping further progress. In this talk, we will follow an alternative made possible by recent innovations in actuators: light homemade robots, of the kind that can bump, fall on, or be lifted by us with no harm. By aiming for less actuators and lower complexity, we end up with morphologies that don't fit exactly the bill of previous ideas. This is a good start to revisit them! We will tour some examples from the open source software and hardware of Upkie, a wheeled biped robot that proudly stands on 3D printed parts and sawed broomsticks (among other mechanical marvels).Stéphane is a roboticist who likes things that balance and walk. He has worked on the locomotion of HRP-4/HRP-2Kai humanoids, ANYmal quadrupeds, and more recently homemade robots like Upkie.#ami #artificialintelligence #robot #mechanical #robotics #technology #homemade #3dprinter #talk
The President of the Italian Republic Sergio Mattarella shook hand with our iCub3 Artificial and Mechanical Intelligence 2022-11-28 | During Made in Inail 2022 we had the pleasure to share our researches with the President of the Italian Republic.
Whole-Body Control and Estimation of Humanoid Robots with Link Flexibility Artificial and Mechanical Intelligence 2022-11-28 | Authors: Giulio Romualdi, Nahuel A. Villa, Stefano Dafarra, Daniele Pucci and Olivier StasseConference: IEEE Humanoids 2022Paper: https://hal.archives-ouvertes.fr/hal-03866027This article presents a whole-body controller for humanoid robots affected by concentrated link flexibility. We characterize the link flexibility by introducing passive joints at the concentration of deflections, which separate the flexible links into two or more rigid bodies. In this way, we extend our robot model to take link deflections into account as underactuated extra degrees of freedom, allowing us to design a whole-body controller capable to anticipate deformations. Since in a real scenario, the deflection is not directly measurable, we present an observer aiming at estimating the flexible joint state, namely position, velocity, and torque, only considering the measured contact force and the state of actuated joint. We validate the overall approach in simulations with the humanoid robot TALOS, whose hip is mechanically flexible due to a localized mechanical weakness. Furthermore, the paper compares the proposed wholebody control strategy with state-of-the-art approaches. Finally, we analyze the performance of the estimator in the case of different values of hip elasticity.
Estimation of Human Base Kinematics using Dynamical Inverse Kinematics and Contact-Aided Lie Group Artificial and Mechanical Intelligence 2022-11-28 | Author: Prashanth Ramadoss, Lorenzo Rapetti, Yeshavi Tirupachuri, Riccardo Grieco, Gianluca Milani, Enrico Valli, Stefano Dafarra, Silvio Traversaro, Daniele PucciFull body motion estimation of a human through wearable sensing technologies is challenging in the absence of position sensors since base kinematics is usually not directly measurable. This paper contributes to the development of a model-based floating base kinematics estimation algorithm using wearable distributed inertial and force-torque sensing. This is done by extending the existing dynamical optimization- based Inverse Kinematics (IK) approach for joint state estimation, in cascade, to include a center of pressure based contact detector and a contact-aided Kalman filter on Lie groups for floating base pose estimation. The proposed method is tested in an experimental scenario where a human equipped with a sensorized suit and shoes performs walking motions. The proposed method is demonstrated to obtain a reliable reconstruction of the whole-body human motion.
Invited talk: Towards mapping the subsurface unknown by Shehryar Khattak, NASA Artificial and Mechanical Intelligence 2022-11-04 | This talk will provide insights into the challenges and opportunities that lie in the field deployment of robots in dark, dirty, and dangerous environments, often characterized by their difficulty to navigate and their adversarial nature towards robot perception due to limited illumination and the presence of obscurants such as dust, fog, and smoke. One representative environment of these challenges is the underground, in the form of subterranean mines, tunnels, caves, and urban passageways. Leveraging these examples, a discussion will be made on the effective use of visual, thermal, depth, and inertial sensors for robot navigation and lessons learned during real-world robotic field deployments, including participation in the recently concluded DARPA Subterranean (SubT) Challenge.
Invited Talk: Outperforming Human Pilots in Autonomy by Dr.Davide Scaramuzza, University of Zurich. Artificial and Mechanical Intelligence 2022-10-12 | We invited Dr. Davide Scaramuzza, from the University of Zurich, to talk about Outperforming Human Pilots in Autonomy.He summarized the latest research in learning deep sensorimotor policies for agile vision-based quadrotor flight. Learning sensorimotor policies represents a holistic approach that is more resilient to noisy sensory observations and imperfect world models. However, training robust policies requires a large amount of data. He will show that simulation data is enough to train policies that transfer to the real world without fine-tuning. They achieved one-shot sim-to-real transfer through the appropriate abstraction of sensory observations and control commands. He showed that these learned policies enable autonomous quadrotors to fly faster and more robustly than before, using only onboard cameras and computation. Applications include acrobatics, high-speed navigation in the wild, and autonomous drone racing.Here the link to the Autonomous Drones made by SuperQuark RAI1.youtube.com/watch?v=fIoydrCbBAMBIO:Davide Scaramuzza is a Professor of Robotics and Perception at the University of Zurich, where he does research at the intersection of robotics, computer vision, and machine learning. His goal is to enable autonomous, agile navigation of micro drones using both standard and neuromorphic event-based cameras. He pioneered autonomous, vision-based navigation of drones, which inspired the navigation algorithm of the NASA Mars helicopter. He has served as a consultant for the United Nations on topics such as disaster response and disarmament, as well as the Fukushima Action Plan on Nuclear Safety. He won many prestigious awards for his research contributions, such as a European-Research-Council Consolidator grant, the IEEE Robotics and Automation Society Early Career Award, an SNF-ERC Starting Grant, a Google Research Award, a Facebook Distinguished Faculty Research Award, two NASA TechBrief Awards, and several paper awards. In 2015, he co-founded Zurich-Eye, today Facebook Zurich, which developed the world-leading virtual-reality headset, Oculus Quest, which sold over 10 million units. Many aspects of his research have been prominently featured in broader media, such as The New York Times, The Economist, Forbes, BBC News, and Discovery Channel.
Invited talk: Deployment of humanoid robots for challenging tasks by Robert Griffin, IHMC Artificial and Mechanical Intelligence 2022-09-22 | Humanoid robots have the incredible potential to do what people do, with the ability to explore both the environments designed for people as well as the unstructured natural world around us. However, we are a long way from having legged robotic systems that can fulfill this potential. Current humanoid robots are much slower and weaker than their biological counterparts, and are typically limited to contacting the world with their feet only. They also require significant operator oversight to accomplish even the simplest tasks. In this talk, I will discuss our recent advances towards humanoid robots that can explore urban environments. This will include the development of the humanoid robot, Nadia, which is a hybrid hydraulic-electric humanoid designed to have speed and power approaching that of a human. I will also discuss our progress in designing a semi-autonomous behavior framework that enables operators to quickly and efficiently direct the robot through both autonomous and tele-operative modes. Lastly, I will highlight some of the necessary algorithm improvements required for high-speed locomotion both over flat ground and uneven terrain, covering both locomotion, perception, and planning.
Invited talk: Motion Planning around Obstacles with Convex Optimization by Tobia Marcucci, MIT Artificial and Mechanical Intelligence 2022-09-17 | From quadrotors delivering packages in urban areas to robot arms moving in confined warehouses, motion planning around obstacles is a core challenge in modern robotics. Planners based on numerical optimization can design trajectories in high-dimensional spaces while taking into account the robot dynamics. However, in the presence of obstacles, these optimizations become nonconvex and very hard to solve, even just locally. Thus, when facing cluttered environments, roboticists typically fall back to sampling-based planners that do not scale equally well to high dimensions and that struggle with continuous differential constraints. In this talk I will present a new framework that enables convex optimization to efficiently and reliably plan optimal trajectories around obstacles. Specifically, I will focus on collision-free motion planning with cost penalties and hard constraints on the shape, the duration, and the velocity of the trajectory. Using optimization techniques that we recently developed for finding shortest paths in graphs of convex sets, we design a practical convex relaxation of the planning problem. This relaxation is typically very tight, to the point that a cheap post-processing of its solution is almost always sufficient to identify a globally-optimal collision-free trajectory. Through numerical and hardware experiments, I will demonstrate that our planner can outperform widely-used sampling-based algorithms and can reliably design trajectories in high-dimensional cluttered environments.
iCub Whole-body jumping Artificial and Mechanical Intelligence 2022-09-07 | Torque and Velocity Controllers to Perform Jumps With a Humanoid Robot: Theory and Implementation on the iCub Robot - IEEE International Conference on Robotics and Automation (ICRA) - 2019
Dynamic Complementarity Conditions & WholeBody Trajectory Optimization for Humanoid Robot Locomotion Artificial and Mechanical Intelligence 2022-08-16 | Dynamic Complementarity Conditions and Whole-Body Trajectory Optimization for Humanoid Robot Locomotion. Paper: ieeexplore.ieee.org/abstract/document/9847574 Preprint: arxiv.org/abs/2207.03198 Repo: github.com/ami-iit/paper_dafarra_2022_tro_dcc-plannerAbstract:The paper presents a planner to generate walking trajectories by using the centroidal dynamics and the full kinematics of a humanoid robot. The interaction between the robot and the walking surface is modeled explicitly via new conditions, the \emph{Dynamic Complementarity Conditions}. The approach does not require a predefined contact sequence and generates the footsteps automatically. We characterize the robot control objective via a set of tasks, and we address it by solving an optimal control problem. We show that it is possible to achieve walking motions automatically by specifying a minimal set of references, such as a constant desired center of mass velocity and a reference point on the ground. Furthermore, we analyze how the contact modelling choices affect the computational time. We validate the approach by generating and testing walking trajectories for the humanoid robot iCub.
Online Control of Humanoid Robot Locomotion: PhD Defense Giulio Romualdi Artificial and Mechanical Intelligence 2022-08-05 | ...
Remote-controlled iCub 3 robot, from Genoa to Venice, thanks to our wearable technologies Artificial and Mechanical Intelligence 2022-07-18 | ...
Modeling and Control of Morphing Covers for the Adaptive Morphology of Humanoid Robots Artificial and Mechanical Intelligence 2022-07-05 | Modeling and Control of Morphing Covers for the Adaptive Morphology of Humanoid RobotsPaper link: ieeexplore.ieee.org/document/9793615arXiv link: arxiv.org/abs/2207.01025 Code link: github.com/ami-iit/paper_bergonti_2022_tro_kinematics-control-morphingcoversLab link: https://ami.iit.it/Abstract: This article takes a step to provide humanoid robots with adaptive morphology abilities. We present a systematic approach for enabling robotic covers to morph their shape, with an overall size fitting the anthropometric dimensions of a humanoid robot. More precisely, we present a cover concept consisting of two main components: a skeleton, which is a repetition of a basic element called node, and a soft membrane, which encloses the cover and deforms with its motion. This article focuses on the cover skeleton and addresses the challenging problems of node design, system modeling, motor positioning, and control design of the morphing system. The cover modeling focuses on kinematics, and a systematic approach for defining the system kinematic constraints is presented. Then, we apply genetic algorithms to find the motor locations so that the morphing cover is fully actuated. Finally, we present control algorithms that allow the cover to morph into a time-varying shape. The entire approach is validated by performing kinematic simulations with four different covers of square dimensions and having 3x3, 4x8, 8x8, and 20x20 nodes, respectively. For each cover, we apply the genetic algorithms to choose the motor locations and perform simulations for tracking a desired shape. The simulation results show that the presented approach ensures the covers to track a desired shape with good tracking performances.0:00 Introduction0:12 Adaptive Morphology1:20 Contributions1:55 Flying Humanoid Robot2:34 Modeling4:37 Actuation Positioning5:22 Control Design5:50 Simulation6:09 Simulation - Scenario 1: Morphing Cover 3x36:39 Simulation - Scenario 2: Morphing Cover 8x87:22 Simulation - Scenario 3: Morphing Cover 20x208:02 Simulation - Scenario 4: Morphing Cover 4x88:57 Conclusion
