Antonio Di Tecco

AI-driven assistance can help the user perform complex teleoperated tasks, introduce autonomous patterns, or adapt the workbench to objects of interest. On the other hand, the level of assistance should be responsive to the user’s response and adapt accordingly to promote a positive and effective experience. Envisaging this final goal, this article investigates whether physiological signals can be used to estimate the user’s performance and response in a teleoperation setup, with and without AI-driven assistance. In more detail, a teleoperated pick-and-place task was performed with or without AI-driven assistance during the grasping phase. A deep-learning algorithm for affordance detection provided assistance, helping participants align the robotic hand with the target object. Physiological and kinematic data were measured and processed by machine learning models to predict the effects of AI assistance on task performance during teleoperation. Results showed that AI-driven assistance, as expected, affected pick-and-place performance. Beyond this, the assistance affected the participant’s fatigue level, which the machine learning models could predict with an average accuracy of 84% based on the physiological response. In addition, the success or failure of the pick-and-place task could be predicted with an average accuracy of 88%. These findings highlight the potential of integrating deep learning with biometric feedback and gesture-based control to create more intuitive and adaptive HRI systems.

Emotions are a key determinant of User Experience (UX). This work investigates the relationship between facial emotions and UX, presenting ALPHA, a predictive system that estimates User Satisfaction based on valence–arousal (VA) signals. The Dataset DT22, a database of 112 participants, is used to evaluate models on the external test set. An identity-free processing pipeline that uses only VA features achieves ∼77% accuracy/weighted F₁-score. When context information is also available—Movie Clip ID and perceived User Quality—the context-aware pipeline increases performance to ∼85–92% in ablation analyses, by using ensembles with majority voting decision rule. This indicates that contextual features provide complementary information to VA. To assess system robustness, Leave-One-Participant-Out (LOPO) and Leave-One-Movieclip-Out (LOMO) protocol schemes, as well as emotion-anchored evaluations (first- and last-emotion), are investigated. The approach is modular and adheres to privacy-by-design principles by operating on identity-free descriptors rather than storing facial frames. Overall, the findings suggest that emotion signals can support adaptive UX and enable dynamic content personalization in applications such as video streaming, customer-experience management, and e-learning. This research presents a practical study on integrating emotion-aware intelligence while respecting user privacy.

This research study proposes a new method to improve robotic teleoperation during grasping tasks. In particular, the target is to reduce the operator's mental burden in grasping tasks, ensuring that the robotic grasping task is simplified for the operator. This simplification aims to improve the task grasping execution time and, in the meantime, increase its grasping success rate. The method is based on an adaptive 3D Virtual Activation Volume (VaV), which acts as a dynamic constraint to support the operator in the grasping task. This volume is centered around the object to grasp and leverages advanced 3D modeling and active control techniques by using machine learning algorithms. In conclusion, the experimental results demonstrate that the proposed method enhances the precision of Human-Robot Interaction (HRI), resulting in a more intuitive teleoperation experience, with a 24% decrease in grasping execution time and a 26.4% increase in grasping success rate.

Predicting user satisfaction in Human–Robot Interaction (HRI) tasks is essential to enhancing system adaptability and user experience. This study addresses this challenge in the context of teleoperated Autonomous Ground Vehicles (AGVs) by introducing the Robot Motion Dataset (RMD). This multimodal dataset integrates motion, control, and haptic feedback signals with user satisfaction scores. Data were collected from 30 participants performing navigation tasks with and without a haptic-actuated glove interface. A set of statistical features was extracted from inertial measurements, control commands, and haptic feedback signals, and the most informative features were selected through a sequential forward selection process. Several machine learning algorithms were trained to classify satisfaction levels, with evaluation performed on both internal and participant-independent external test sets. The best performance was achieved by a Weighted k-Nearest Neighbors classifier, reaching an accuracy above 80% in both experimental conditions. The results demonstrate the feasibility of predicting user satisfaction from multimodal sensor data in real time, highlighting the potential of the proposed framework for adaptive HRI systems.

This research study investigates the impact of a virtual dashboard on the quality of task execution in robotic teleoperation. More specifically, this study investigates how a virtual dashboard improves user awareness and grasp precision in a teleoperated pick-and-place task by providing users with critical information in real-time. An experiment was conducted with 30 participants in a robotic teleoperated task to measure their task performance in two different experimental conditions: a control group used conventional interfaces, and an experimental group utilized the virtual dashboard with additional information. Research findings indicate that integrating a virtual dashboard improves grasping accuracy, reduces user fatigue, and speeds up task completion, thereby improving task effectiveness and the quality of the experience.

A known issue in teleoperation concerns dexterity in grasping objects, which can be severely limited if not properly coupled with perfect stereoscopic vision and force/tactile feedback. This work presents the development and testing of a subsystem used for a virtual dashboard within the Sully teleoperation system, with hardware and software components. It is used to capture images, gather object information, and display a virtual dashboard that provides the operator with information about the distance and reachability of objects, to support stereoscopic vision visualized through a virtual reality headset. The hardware consists of two front cameras, used as eyes for the teleoperated robot, and two Time-of-Flight (ToF) sensors to estimate the distance between the teleoperated robot and the Object of Interest (OoI). The software processes images captured by cameras to identify the objects, detects the OoI, and estimates its distance from the robot using the measurements of ToF sensors. Performance results showed an optimal trade-off between image quality, frame rate, and use experience. Indeed, a mean accuracy of 88% was achieved in identifying and detecting OoIs, and an average mean absolute percentage error (MAPE) of 1.40% to estimate the distance from the robot to the OoI at 120 cm.

With the explosion of online video consumption, assessing and anticipating how users will evaluate the content they watch has become increasingly important. Traditional methods based on explicit user feedback are often limited in their ability to do this, as they can be time-consuming and expensive to collect. This study explores techniques to predict users’ ratings about a video’s ability to evoke emotions through emotional signals. In particular, it is proposed a method of emotional analysis that uses valence and arousal data as key signals for predicting user ratings through systems that use machine-learning techniques. Hence, an experiment in the wild involved 112 participants who completed questionnaires to create a dataset of emotional data and video quality ratings to train different intelligent systems. The best system comprised a Medium Gaussian Support Vector Machine (SVM) classifier that detected users’ ratings between Ineffective and Effective based on valence and arousal features as input, achieving an accuracy higher than 87%. The result demonstrated that it is possible to predict users’ ratings on the ability of the movie to elicit emotion, using users’ emotional states in terms of valence and arousal. The system has several advantages, such as eliminating the need for user reports, predicting user ratings in real-time more quickly and dynamically, and utilizing only the initial emotional state to predict users’ ratings. In addition, it has potential applications in advertising, education, and entertainment fields. Advertisers could better understand how consumers perceive their products and create more effective advertising campaigns; educational institutions could develop more engaging and effective learning materials; entertainment providers could create more popular and successful content.

Stress can be defined as a state of anxiety (or mental tension) caused by a particular situation. Everybody experiences stress to some level, but how we respond to stress significantly affects our well-being. Various events generate anxiety that leads to stress. For example, not having enough time to complete a task or being late are situations where anxiety (and stress) depends on a temporal factor: the scarcity of time. But people also slide into anxiety as they live in a condition that causes them to be tense, independently of time. The studies eliciting anxiety in laboratory settings have less widely considered this variant. This paper presents a proof of concept (PoC) that investigated the possibility of stimulating anxiety without time pressure through a purposely edited horror movie trailer, giving new insights into the emotional experiences evoked by controlled audiovisual stimuli. The PoC comprised an AI-based classifier to detect a person’s emotion among anxiety , relaxation , and none of the two based on the galvanic skin response (GSR), photoplethysmogram (PPG), and heart rate (HR), achieving an accuracy higher than 96%. Key application areas include media and marketing, and psychology. Media producers could improve their content to capture the audience better; psychologists could create tailored exposure experiences to promote gradual desensitization to stress triggers.