Stefano Dafarra
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Stefano Dafarra has a total of 98 co-authors. Here are the top co-authors by number of shared publications:
- Daniele Pucci
Artificial and Mechanical Intelligence, Istituto Italiano di Tecnologia, Genoa, Italy
13 shared publications - Giulio Romualdi
Artificial and Mechanical Intelligence, Istituto Italiano di Tecnologia, Genoa, Italy
8 shared publications - Silvio Traversaro
Artificial and Mechanical Intelligence, Istituto Italiano di Tecnologia, Genoa, Italy
5 shared publications Ines Sorrentino
Istituto Italiano di Tecnologia
4 shared publications- Lorenzo Rapetti
Artificial and Mechanical Intelligence (AMI), Istituto Italiano di Tecnologia (IIT), Genoa, Italy
3 shared publications Mohamed Elobaid
Istituto Italiano di Tecnologia
3 shared publications
Latest Publications
- Open Access
- 1 January 2026
- Published by Wiley in Human Behavior and Emerging Technologies
- Vol. 2026 (1)
Theta Frequency Band in the EEG Signal as Neurocognitive Indicator of Cognitive Load for Collaborative Tasks Involving Robot Teleoperation
The aim of the present study was to identify metrics of cognitive effort during robot teleoperation in a social context of joint collaborative task performed with another human user. Robotics teleoperation enables control of artificial systems in remote environments but is often cognitively challenging. Measuring human cognitive effort during teleoperation has largely relied on subjective measures like questionnaires. To evaluate the quality of teleoperated systems, it is critical to investigate objective measures of cognitive effort directly in natural contexts, especially during social interactions with individuals encountering the teleoperated robot. In two separate experiments, we recorded—with dual EEG—brain activity of participants performing a cooperative task. In the first experiment, participants were asked to perform a cooperative task in a direct human–human interaction setup. In the second experiment, the interaction was mediated by the iCub robot, teleoperated by one of the participants using a VR headset (human–robot–human interaction). The EEG data revealed an increase of power in the theta (4–7 Hz) frequency band in the robot‐mediated interaction, relative to human–human interaction, likely reflecting the additional cognitive load required by robot teleoperation. This demonstrates that using EEG in teleoperation might provide a reliable metric of the quality of the system and a way to tap onto cognitive mechanisms underlying the user′s experience in teleoperation.
- 1 January 2026
- Published by Wiley in Human Behavior and Emerging Technologies
- Vol. 2026 (1)
Open AccessTheta Frequency Band in the EEG Signal as Neurocognitive Indicator of Cognitive Load for Collaborative Tasks Involving Robot Teleoperation
The aim of the present study was to identify metrics of cognitive effort during robot teleoperation in a social context of joint collaborative task performed with another human user. Robotics teleoperation enables control of artificial systems in remote environments but is often cognitively challenging. Measuring human cognitive effort during teleoperation has largely relied on subjective measures like questionnaires. To evaluate the quality of teleoperated systems, it is critical to investigate objective measures of cognitive effort directly in natural contexts, especially during social interactions with individuals encountering the teleoperated robot. In two separate experiments, we recorded—with dual EEG—brain activity of participants performing a cooperative task. In the first experiment, participants were asked to perform a cooperative task in a direct human–human interaction setup. In the second experiment, the interaction was mediated by the iCub robot, teleoperated by one of the participants using a VR headset (human–robot–human interaction). The EEG data revealed an increase of power in the theta (4–7 Hz) frequency band in the robot‐mediated interaction, relative to human–human interaction, likely reflecting the additional cognitive load required by robot teleoperation. This demonstrates that using EEG in teleoperation might provide a reliable metric of the quality of the system and a way to tap onto cognitive mechanisms underlying the user′s experience in teleoperation.
- 24 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 10 (11) , 11674-11681
Contact-Implicit Whole-Body Trajectory Optimization for Dynamic Humanoid Locomotion on Parameterizable Uneven Terrain
This letter presents a trajectory planner for humanoid robots to perform dynamic motions on complex, parameterizable terrains using a contact-implicit whole-body optimization framework. The planner generates feasible movements without predefined contact sequences, considering the robot full dynamics and kinematics. To plan bipedal locomotion on steps, ramps, and other terrains, we propose a mathematical model to approximate ground geometry with a smooth surface. The rigid interaction between the robot and the ground is modeled via novel contact constraints, extending the domain of applicability from flat and horizontal surfaces to any arbitrary smooth surface. Effectiveness is demonstrated using nonlinear programming to plan motions on various terrain topologies. We also explore dynamic trajectory generation by embedding high-level requirements in the cost function and the initial guess. Preliminary feasibility tests were conducted over non-coplanar terrains, both in simulation and on the real robot ErgoCub.
- 24 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 10 (11) , 11674-11681
Contact-Implicit Whole-Body Trajectory Optimization for Dynamic Humanoid Locomotion on Parameterizable Uneven Terrain
This letter presents a trajectory planner for humanoid robots to perform dynamic motions on complex, parameterizable terrains using a contact-implicit whole-body optimization framework. The planner generates feasible movements without predefined contact sequences, considering the robot full dynamics and kinematics. To plan bipedal locomotion on steps, ramps, and other terrains, we propose a mathematical model to approximate ground geometry with a smooth surface. The rigid interaction between the robot and the ground is modeled via novel contact constraints, extending the domain of applicability from flat and horizontal surfaces to any arbitrary smooth surface. Effectiveness is demonstrated using nonlinear programming to plan motions on various terrain topologies. We also explore dynamic trajectory generation by embedding high-level requirements in the cost function and the initial guess. Preliminary feasibility tests were conducted over non-coplanar terrains, both in simulation and on the real robot ErgoCub.
- Open Access
- 16 April 2025
- Published by Springer Nature in Research Square
Towards a Shared Embodied Intelligence of Humanoid Robots: Optimization, Development, and Testing of the Human-Aware ergoCub Robot
Collaboration is an inherent aspect of human behavior, enabling tasks that are otherwise impossible for individuals alone. This collaborative ability stems from the capacity of individuals to coordinate their actions by leveraging internal representations of their companions—a concept known as shared intelligence. Additionally, humans are characterized by physical bodies and cognitive abilities that are optimized in response to their environment, a phenomenon referred to as embodied cognition. In this paper, we propose an architecture that combines shared intelligence and embodied cognition to enable robots to physically collaborate with humans, where robot hardware and control are optimized for human metrics, using representations of the human body and motion intelligence. By doing so, our ultimate aim is to design humanoid robots that possess a degree of shared embodied intelligence. More specifically, the proposed architecture optimizes the robot’s hardware and physical intelligence parameters with respect to human ergonomic metrics. We achieve this by deriving models of human-robot interaction parameterized with respect to the robot’s hardware configurations and by incorporating human models into the robot’s physical intelligence. As a tangible outcome of the proposed architecture, we present the humanoid robot ergoCub, whose body and physical intelligence have been optimized for performing collaborative actions with humans. Our approach yields concrete results in designing humanoid robots that prioritize human ergonomics at both the hardware and physical intelligence levels. The potential applications of this approach span diverse fields, including industrial and assistive robotics.
- 16 April 2025
- Published by Springer Nature in Research Square
Open AccessTowards a Shared Embodied Intelligence of Humanoid Robots: Optimization, Development, and Testing of the Human-Aware ergoCub Robot
Collaboration is an inherent aspect of human behavior, enabling tasks that are otherwise impossible for individuals alone. This collaborative ability stems from the capacity of individuals to coordinate their actions by leveraging internal representations of their companions—a concept known as shared intelligence. Additionally, humans are characterized by physical bodies and cognitive abilities that are optimized in response to their environment, a phenomenon referred to as embodied cognition. In this paper, we propose an architecture that combines shared intelligence and embodied cognition to enable robots to physically collaborate with humans, where robot hardware and control are optimized for human metrics, using representations of the human body and motion intelligence. By doing so, our ultimate aim is to design humanoid robots that possess a degree of shared embodied intelligence. More specifically, the proposed architecture optimizes the robot’s hardware and physical intelligence parameters with respect to human ergonomic metrics. We achieve this by deriving models of human-robot interaction parameterized with respect to the robot’s hardware configurations and by incorporating human models into the robot’s physical intelligence. As a tangible outcome of the proposed architecture, we present the humanoid robot ergoCub, whose body and physical intelligence have been optimized for performing collaborative actions with humans. Our approach yields concrete results in designing humanoid robots that prioritize human ergonomics at both the hardware and physical intelligence levels. The potential applications of this approach span diverse fields, including industrial and assistive robotics.
- Open Access
- 10 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 10 (5) , 4348-4355
Maximising Tolerance to Disturbances via Combined Control-Actuation Optimisation for Robust Humanoid Robot Walking
Combined optimisation of various robot subsystems as a co-design problem has been shown to identify performant robots. However, classical optimisation methods result in point-optimum solutions that may not ensure robust performance and physical feasibility, i.e., the existence of components with specifications matching the computed optimum value. To address this problem, we present a set-based robust co-design optimisation strategy to maximise disturbance tolerance. Instead of identifying a single point-optimum solution, a so-called solution space evaluates the combination of the largest design space that delivers the necessary performance while being robust to the largest set of disturbances. The utility of the proposed approach is demonstrated via a computational design study of the ergoCub robot. This study focuses on the robots' walking performance, illustrating (1) improvement in task success considering at least 3 times larger magnitudes of disturbances, (2) identifying a set instead of a point-solution in the design-disturbances space, and (3) improving standardisation of the joint actuation design.
- 10 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 10 (5) , 4348-4355
Open AccessMaximising Tolerance to Disturbances via Combined Control-Actuation Optimisation for Robust Humanoid Robot Walking
Combined optimisation of various robot subsystems as a co-design problem has been shown to identify performant robots. However, classical optimisation methods result in point-optimum solutions that may not ensure robust performance and physical feasibility, i.e., the existence of components with specifications matching the computed optimum value. To address this problem, we present a set-based robust co-design optimisation strategy to maximise disturbance tolerance. Instead of identifying a single point-optimum solution, a so-called solution space evaluates the combination of the largest design space that delivers the necessary performance while being robust to the largest set of disturbances. The utility of the proposed approach is demonstrated via a computational design study of the ergoCub robot. This study focuses on the robots' walking performance, illustrating (1) improvement in task success considering at least 3 times larger magnitudes of disturbances, (2) identifying a set instead of a point-solution in the design-disturbances space, and (3) improving standardisation of the joint actuation design.
- 30 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 10 (3) , 2806-2813
Adaptive Non-Linear Centroidal MPC With Stability Guarantees for Robust Locomotion of Legged Robots
Nonlinear model predictive locomotion controllers based on the reduced centroidal dynamics are nowadays ubiquitous in legged robots. These schemes, even if they assume an inherent simplification of the robot's dynamics, were shown to endow robots with a step-adjustment capability in reaction to small pushes, and in the case of uncertain parameters - as unknown payloads - they were shown to provide some “practical”, albeit limited, robustness. In this work, we provide rigorous certificates of their closed-loop stability via reformulating the online centroidal MPC controller. This is achieved thanks to a systematic procedure inspired by the machinery of adaptive control, together with ideas coming from Control Lyapunov Functions. Our reformulation, in addition, provides robustness for a class of unmeasured constant disturbances. To demonstrate the generality of our approach, we validated our formulation on a new generation of humanoid robots - the $\text{56.7 kg}$ ergoCub, as well as on the commercially available $\text{21 kg}$ quadruped robot Aliengo.
- 30 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 10 (3) , 2806-2813
Adaptive Non-Linear Centroidal MPC With Stability Guarantees for Robust Locomotion of Legged Robots
Nonlinear model predictive locomotion controllers based on the reduced centroidal dynamics are nowadays ubiquitous in legged robots. These schemes, even if they assume an inherent simplification of the robot's dynamics, were shown to endow robots with a step-adjustment capability in reaction to small pushes, and in the case of uncertain parameters - as unknown payloads - they were shown to provide some “practical”, albeit limited, robustness. In this work, we provide rigorous certificates of their closed-loop stability via reformulating the online centroidal MPC controller. This is achieved thanks to a systematic procedure inspired by the machinery of adaptive control, together with ideas coming from Control Lyapunov Functions. Our reformulation, in addition, provides robustness for a class of unmeasured constant disturbances. To demonstrate the generality of our approach, we validated our formulation on a new generation of humanoid robots - the $\text{56.7 kg}$ ergoCub, as well as on the commercially available $\text{21 kg}$ quadruped robot Aliengo.
- 11 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 9 (12) , 11617-11624
XBG: End-to-End Imitation Learning for Autonomous Behaviour in Human-Robot Interaction and Collaboration
This paper presents XBG (eXteroceptive Behaviour Generation), a multimodal end-to-end Imitation Learning (IL) system for whole-body autonomous humanoid robots used in real-world Human-Robot Interaction (HRI) scenarios. The main contribution is an architecture for learning HRI behaviours using a data-driven approach. A diverse dataset is collected via teleoperation, covering multiple HRI scenarios, such as handshaking, handwaving, payload reception, walking, and walking with a payload. After synchronizing, filtering, and transforming the data, we show how to train the presented Deep Neural Networks (DNN), integrating exteroceptive and proprioceptive information to help the robot understand both its environment and its actions. The robot takes in sequences of images (RGB and depth) and joints state information to react accordingly. By fusing multimodal signals over time, the model enables autonomous capabilities in a robotic platform. The models are evaluated based on the success rates in the mentioned HRI scenarios and they are deployed on the ergoCub humanoid robot. XBG achieves success rates between 60% and 100% even when tested in unseen environments.
- 11 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 9 (12) , 11617-11624
XBG: End-to-End Imitation Learning for Autonomous Behaviour in Human-Robot Interaction and Collaboration
This paper presents XBG (eXteroceptive Behaviour Generation), a multimodal end-to-end Imitation Learning (IL) system for whole-body autonomous humanoid robots used in real-world Human-Robot Interaction (HRI) scenarios. The main contribution is an architecture for learning HRI behaviours using a data-driven approach. A diverse dataset is collected via teleoperation, covering multiple HRI scenarios, such as handshaking, handwaving, payload reception, walking, and walking with a payload. After synchronizing, filtering, and transforming the data, we show how to train the presented Deep Neural Networks (DNN), integrating exteroceptive and proprioceptive information to help the robot understand both its environment and its actions. The robot takes in sequences of images (RGB and depth) and joints state information to react accordingly. By fusing multimodal signals over time, the model enables autonomous capabilities in a robotic platform. The models are evaluated based on the success rates in the mentioned HRI scenarios and they are deployed on the ergoCub humanoid robot. XBG achieves success rates between 60% and 100% even when tested in unseen environments.
- 30 January 2024
- Published by Springer Nature in International Journal of Social Robotics
- Vol. 17 (3) , 473-504
Analysis and Perspectives on the ANA Avatar XPRIZE Competition
- 30 January 2024
- Published by Springer Nature in International Journal of Social Robotics
- Vol. 17 (3) , 473-504
Analysis and Perspectives on the ANA Avatar XPRIZE Competition
- Open Access
iCub3 Avatar System: Enabling Remote Fully-Immersive Embodiment of Humanoid Robots
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.
Open AccessiCub3 Avatar System: Enabling Remote Fully-Immersive Embodiment of Humanoid Robots
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.
- Top Cited Papers
- 24 January 2024
- Published by American Association for the Advancement of Science (AAAS) in Science Robotics
- Vol. 9 (86) , eadh3834
iCub3 avatar system: Enabling remote fully immersive embodiment of humanoid robots
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. More precisely, the paper makes two contributions: First, we present the humanoid iCub3 as a robotic avatar that integrates the latest significant improvements after about 15 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 facial expressions with comprehensive sensory feedback including visual, auditory, haptic, weight, and touch modalities. We validated the system by implementing several avatar architecture instances, each tailored to specific requirements. First, we evaluated the optimized architecture for verbal, nonverbal, 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 kilometers away—thus allowing the operator to visit the Italian art exhibition remotely. Second, we evaluated the optimized 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 operated in Rimini—about 300 kilometers away—interacting with a recipient who entrusted the avatar with 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 All Nippon Airways (ANA) Avatar XPrize competition.
- 24 January 2024
- Published by American Association for the Advancement of Science (AAAS) in Science Robotics
- Vol. 9 (86) , eadh3834
Top Cited PapersiCub3 avatar system: Enabling remote fully immersive embodiment of humanoid robots
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. More precisely, the paper makes two contributions: First, we present the humanoid iCub3 as a robotic avatar that integrates the latest significant improvements after about 15 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 facial expressions with comprehensive sensory feedback including visual, auditory, haptic, weight, and touch modalities. We validated the system by implementing several avatar architecture instances, each tailored to specific requirements. First, we evaluated the optimized architecture for verbal, nonverbal, 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 kilometers away—thus allowing the operator to visit the Italian art exhibition remotely. Second, we evaluated the optimized 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 operated in Rimini—about 300 kilometers away—interacting with a recipient who entrusted the avatar with 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 All Nippon Airways (ANA) Avatar XPrize competition.
- Open Access
- 2 August 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Robotics
- Vol. 38 (6) , 3414-3433
Dynamic Complementarity Conditions and Whole-Body Trajectory Optimization for Humanoid Robot Locomotion
This article 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 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 modeling choices affect the computational time. We validate the approach by generating and testing walking trajectories for the humanoid robot iCub.
- 2 August 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Robotics
- Vol. 38 (6) , 3414-3433
Open AccessDynamic Complementarity Conditions and Whole-Body Trajectory Optimization for Humanoid Robot Locomotion
This article 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 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 modeling choices affect the computational time. We validate the approach by generating and testing walking trajectories for the humanoid robot iCub.