A5050576327- Robot Manipulation and Learning
- Muscle activation and electromyography studies
- Motor Control and Adaptation
- Advanced Neural Network Applications
- Soft Robotics and Applications
- Stroke Rehabilitation and Recovery
- Multimodal Machine Learning Applications
- Tactile and Sensory Interactions
- Human Pose and Action Recognition
- EEG and Brain-Computer Interfaces
- Social Robot Interaction and HRI
- Anomaly Detection Techniques and Applications
- Reinforcement Learning in Robotics
- Adversarial Robustness in Machine Learning
- Prosthetics and Rehabilitation Robotics
- Robotic Mechanisms and Dynamics
- Action Observation and Synchronization
- Video Analysis and Summarization
- Botulinum Toxin and Related Neurological Disorders
- Robotic Locomotion and Control
- Domain Adaptation and Few-Shot Learning
- Cerebral Palsy and Movement Disorders
- Teleoperation and Haptic Systems
- COVID-19 diagnosis using AI
- Gaze Tracking and Assistive Technology
Polytechnic University of Turin
2021-2025
University of Pisa
2017-2024
Piaggio (Italy)
2017-2022
Italian Institute of Technology
2017-2021
Centro Universitário Plínio Leite
2021
Soft hands are robotic systems that embed compliant elements in their mechanical design.This enables an effective adaptation with the items and environment, ultimately, increase grasping performance.These come clear advantages terms of ease-to-use robustness if compared classic rigid hands, when operated by a human.However, potential for autonomous is still largely unexplored, due to lack suitable control strategies.To address this issue, letter, we propose approach enable soft autonomously...
In the last decade, most research in Machine Learning contributed to improvement of existing models, with aim increasing performance neural networks for solution a variety different tasks. However, such advancements often come at cost an increase model memory and computational requirements. This represents significant limitation deployability output realistic settings, where cost, energy consumption, complexity framework play crucial role. To solve this issue, designer should search models...
Humans are able to intuitively exploit the object shape and environmental constraints drive hand in order achieve stable grasps perform dexterous manipulations. Despite vast range of kinematic strategies employed by humans, this work we consider test hypothesis that such ability can be described terms a synergistic behavior generation postures, i.e. using reduced set commonly used patterns. To do performed experiments on six subjects, who were asked grasp objects from flat surface. We...
Abstract Background Shedding light on the neuroscientific mechanisms of human upper limb motor control, in both healthy and disease conditions (e.g., after a stroke), can help to devise effective tools for quantitative evaluation impaired conditions, properly inform rehabilitative process. Furthermore, design control mechatronic devices also benefit from such outcomes, with important implications assistive rehabilitation robotics advanced human-machine interaction. To reach these goals, we...
The rich variety of human upper limb movements requires an extraordinary coordination different joints according to specific spatio-temporal patterns. However, unvealing these motor schemes is a challenging task. Principal components have been often used for analogous purposes, but such approach relies on hypothesis temporal uncorrelation poses in time. To overcome limitations, this work we leverage functional Component Analysis (fPCA). We carried out experiments with 7 sbjects performing...
Abstract Background Human-likeliness of robot movements is a key component to enable safe and effective human-robot interaction, since it contributes increase acceptance motion predictability robots that have closely interact with people, e.g. for assistance rehabilitation purposes. Several parameters been used quantify how much behaves like human, which encompass aspects related both the appearance motion. The latter point fundamental allow operator interpret robotic actions, plan...
Blindness represents one of the major disabling societal causes, impacting life visually impaired people and their families. For what concerns access to written information, main tools used by blind is traditional Braille code. This reason why in recent years, there has been a technological effort develop refreshable devices. These consist multiple physical dots that dynamically change configuration reproduce different sequences letters Although promising, these approaches have many...
Deep Neural Networks (DNNs) enable a wide series of technological advancements, ranging from clinical imaging, to predictive industrial maintenance and autonomous driving. However, recent findings indicate that transient hardware faults may corrupt the models prediction dramatically. For instance, radiation-induced misprediction probability can be so high impede safe deployment DNNs at scale, urging need for efficient effective hardening solutions. In this work, we propose tackle reliability...
Deep Neural Networks (DNNs) have revolutionized several fields, including safety- and mission-critical applications, such as autonomous driving space exploration. However, recent studies highlighted that transient hardware faults can corrupt the model's output, leading to high misprediction probabilities. Since traditional reliability strategies, based on modular hardware, software replications, or matrix multiplication checksum impose a overhead, there is pressing need for efficient...
Our comprehension of video streams depicting human activities is naturally multifaceted: in just a few moments, we can grasp what happening, identify the relevance and interactions objects scene, forecast will happen soon, everything all at once. To endow autonomous systems with such holistic perception, learning how to correlate concepts, abstract knowledge across diverse tasks, leverage tasks synergies when novel skills essential. A significant step this direction EgoPack, unified...
Recent functional magnetic resonance imaging (fMRI) studies have identified specific neural patterns related to three different categories of movements: intransitive (i.e., meaningful gestures that do not include the use objects), transitive actions involving an object), and tool-mediated a tool interact with object). However, fMRI intrinsically limits exploitation these results in real scenario, such as brain-machine interface. In this paper, we propose new approach automatically predict...
In this paper, we present a novel approach to dynamically describe human upper limb trajectories, addressing the question on whether and which extent synergistic multi-joint behavior is observed preserved over time evolution across subjects. To goal, performed experiments collect joint angle trajectories organized them in dataset of daily living tasks. We then characterized poses at each frame through technique that named repeated-principal component analysis (R-PCA). found that, although...
Recently, the method of choice to exploit robot dynamics for efficient walking is numerical optimization (NO). The main drawback in NO computational complexity, which strongly affects time demand solution. Several strategies can be used make more treatable and efficiently describe solution set. In this letter, we present an algorithm encode effective references, generated offline via optimization, extracting a limited number principal components using them as basis optimal motions. By...
Postural hand synergies or eigenpostures are joint angle covariation patterns observed in common grasping tasks. A typical definition associates the geometry of synergy vectors and their hierarchy (relative statistical weight) with principal component analysis an experimental covariance matrix. In a reduced complexity representation, accuracy posture reconstruction is incrementally improved as number increased according to hierarchy. this work, we explore whether how incrementality extend...
Recently, the avenue of adaptable, soft robotic hands has opened simplified opportunities to grasp different items; however, potential end effectors (SEEs) is still largely unexplored, especially in human-robot interaction. In this paper, we propose, for first time, a simple touch-based approach endow SEE with autonomous sensory-motor primitives, response an item passed robot by human (human-to-robot handover). We capitalize on inspiration and minimalistic sensing, while hand adaptability...
The accurate assessment of upper limb motion impairment induced by stroke - which represents one the primary causes disability world-wide is first step to successfully monitor and guide patients' recovery. As today, majority procedures relies on clinical scales, are mostly based ordinal scaling, operator-dependent, subject floor ceiling effects. In this work, we intend overcome these limitations proposing a novel approach analytically evaluate level pathological movement coupling,...
Objective.Brain-computer interfaces (BCIs) exploit computational features from brain signals to perform a given task. Despite recent neurophysiology and clinical findings indicating the crucial role of functional interplay between cardiovascular dynamics in locomotion, heartbeat information remains be included common BCI systems. In this study, we multidimensional directional electroencephalographic spectra classify upper limb movements into three classes.Approach.We gathered data 26 healthy...
An effective robotic wrist represents a key enabling element in manipulation, especially prosthetics. In this paper, we propose an under-actuated system, which is also adaptable and allows to implement different under-actuation schemes. Our approach leverages upon the idea of soft synergies — particular design method adaptive as it derives from field robot hand design. First introduce principle its implementation function configurable test bench prototype, can be used demonstrate feasibility...
This paper tackles the challenge of predicting grasp failures in soft hands before they happen, by combining deep learning with a sensing strategy based on distributed Inertial Measurement Units. We propose two neural architectures, which we implemented and tested an articulated hand - Pisa/IIT SoftHand continuously deformable RBO Hand. The first architecture (Classifier) implements a-posteriori detection failure event, serving as test-bench to assess possibility extracting information from...
The need for users' safety and technology accept-ability has incredibly increased with the deployment of co-bots physically interacting humans in industrial settings, people assistance. A well-studied approach to meet these requirements is ensure human-like robot motions. Classic solutions anthropomorphic movement generation usually rely on optimization procedures, which build upon hypotheses devised from neuroscientific literature, or capitalize learning methods. However, approaches come...
Physical interaction of robots with their environment is a challenging problem because the exchanged forces. Hybrid position/force control schemes often exhibit problems during contact phase, whereas impedance appears to be more simple and reliable, especially when shaped energetically passive. Even if recent technologies enable shaping robot, how best plan parameters for task execution remains an open question. In this paper we present optimization-based approach not only robot motion but...
In recent years, the spread of data-driven approaches for robotic grasp synthesis has come with increasing need reliable datasets, which can be built e.g. through video labelling. To this goal, it is important to define suitable rules characterize main human types, easily identifying them in streams. work, we present a novel taxonomy that builds upon related state art, but specifically thought It focuses on interaction hand environment and accounts pre-contact phases, bi-manual grasps as...