United States

In this paper, we build upon two major recent developments in the field, Diffusion Policies for visuomotor manipulation and large pre-trained multimodal foundational models to obtain a robotic skill learning system. The system can obtain new skills via the behavioral cloning approach of visuomotor diffusion policies given teleoperated demonstrations. Foundational models are being used to perform skill selection given the user&#39;s prompt in natural language. Before executing a skill the foundational model performs a precondition check given an observation of the workspace. We compare the performance of different foundational models to this end and give a detailed experimental evaluation of the skills taught by the user in simulation and the real world. Finally, we showcase the combined system on a challenging food serving scenario in the real world. Videos of all experimental executions, as well as the process of teaching new skills in simulation and the real world, are available on the project&#39;s website.

IEEE RO-MAN 2024

A Robotic Skill Learning System Built Upon Diffusion Policies and Foundation Models

foundational models

visuomotor diffusion policies

robotic skill learning

We are pleased to announce the Fortieth AAAI Conference on Artificial Intelligence (AAAI-26), which will be held in Singapore EXPO from January 20 to January 27, 2026.

The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines. AAAI-26 will feature technical paper presentations, special tracks, invited speakers, workshops, tutorials, poster sessions, senior member presentations, competitions, and exhibit programs, and a range of other activities to be announced.

To access this event page, you need to log in with the **email address you registered with**. Access credentials will be sent to your email from Underline - subject line "Welcome to AAAI 2026". Please be sure to check your spam email folder if you do not see an email confirmation right away.

Please log in

To access this event page, you are required to register.
Please complete your registration to continue.

We recommend reading [**the registration information**](https://aaai.org/conference/aaai/aaai-26/registration/) first.

**Online Registration Form**: https://aaai.getregistered.net/conference-2026 

Registration Required

AAAI 2026

We are pleased to announce the Fortieth AAAI Conference on Artificial Intelligence (AAAI-26), which will be held in Singapore EXPO from January 20 to January 27, 2026.

In this paper, we build upon two major recent developments in the field, Diffusion Policies for visuomotor manipulation and large pre-trained multimodal foundational models to obtain a robotic skill learning system. The system can obtain new skills via the behavioral cloning approach of visuomotor diffusion policies given teleoperated demonstrations. Foundational models are being used to perform skill selection given the user's prompt in natural language. Before executing a skill the foundational model performs a precondition check given an observation of the workspace. We compare the performance of different foundational models to this end and give a detailed experimental evaluation of the skills taught by the user in simulation and the real world. Finally, we showcase the combined system on a challenging food serving scenario in the real world. Videos of all experimental executions, as well as the process of teaching new skills in simulation and the real world, are available on the project's website.

technical paper

[![](https://assets.underline.io/markdown_image/1/image/f25361a85b7c83a7476c0b3c1c0bb1b2.png)](https://drive.google.com/file/d/18jcAcciCBQHj6R07Dbpm2HCOBEIwAOCU/view?usp=sharing)


Dear Colleagues and Friends,

The 33rd IEEE International Conference on Robot and Human Interactive Communication, IEEE RO-MAN 2024, was held at Pasadena Convention Center in Pasadena, USA.

The theme of the IEEE RO-MAN 2024 was **“Embracing Human-Centered HRI”**.

With the continuous technological effort of understanding human needs, creating human-desired robotic systems, and developing real human user experiences, we are marching together towards building a promising and excellent human quality of life, all for our wellness and our next generation.

The conference provided a forum for researchers and practitioners to discuss a variety of topics related to the broad theme of Robot and Human Interactive Communication involving theories, methodologies, technologies, empirical and experimental studies, and generated more insights for the future of Robot and Human Interactive Communication!

We hope you continue to enjoy the post-conference content on Underline!

You need to log in with the email address you registered with. Access credentials have been sent to your email. 

Please be sure to check your spam and other email folders if you do not see an email confirmation right away.

Please log in to explore this event.

It looks like you are not registered for this event. 

A registration is required to access this event page. 



Registration Is Required

The 33rd IEEE International Conference on Robot and Human Interactive Communication was held in Pasadena, California, USA. The theme of the IEEE RO-MAN 2024 was “Embracing Human-Centered HRI”.

For social robots to succeed in places such as homes, they must learn new skills from various people and act in a manner desirable to different users. We introduce a novel biologically inspired approach for robot learning through program-level imitation, inspired by the way primates, including humans, understand and perform complex actions. Our approach enables robots to discover the hierarchical structure of tasks by identifying sequential regularities and sub-goals from diverse human demonstrations. To do so, human-provided demonstrations, which can be obtained by a robot through different modalities (such as kinesthetic teaching, behavioural observation, and verbal instruction), are processed by an algorithm that discovers multiple possibilities for arranging observed sub-goals to achieve a final goal. Prior to acting, the available sequences are evaluated based on user-defined criteria, through mental simulation of the task by the robot, to find the optimal sequence of actions. As a proof-of-concept, we implemented our system on an iCub humanoid robot and present here how our method allowed the robot to adapt its action sequences for task execution when starting the task from different states, incorporating user preference for finishing the task as fast as possible. Our envisaged system is meant to accommodate variations in human teaching styles and is expected to help a robot perform tasks with greater flexibility and efficiency. This work contributes by proposing a framework for robots to learn from humans at an abstract level, opening the way to more adaptable and intelligent robotic assistants in everyday tasks.

A Biologically Inspired Program-Level Imitation Approach for Robots: Proof-Of-Concept

Dialogs between humans and social robots are often fact-oriented and not very mood-dependent, at least on the part of the robot. This raises the question of what influence the tonality of the utterances has on the perceived personality of the robot and how this influences the user's further expectations. One element that is not necessarily expected from robots in conversation is humor, especially the use of irony and sarcasm. We examined the effect of a conversation with cheerfully ironic or sullenly sarcastic remarks by the robot in contrast to utterances in a more seriously neutral tone. To measure these effects in terms of perceived warmth, competence and possible discomfort in conversation we used the Robotic Social Attributes Scale (RoSAS) on 20 participants with an age ranging from 26 to 58. In addition, we asked the participants for their assessment of the suitability of the robot's dialog style for different areas of application and which dialog style they personally liked best. The dialog parts of the robot were generated with ChatGPT using suitable prompts. Our results showed that the perceived warmth gained highest rates when when the robot hat a cheerfully ironic tone. Discomfort was rated significantly higher when a sullenly sarcastic tone was used by the robot. Perceived competence seemed to be slightly negatively influenced in the sarcastic condition. The results proved that tonality appears to be a very relevant design element for a successful interaction between humans and robots, but must be used carefully to achieve the desired goal.

Introducing a Note of Levity to Human-Robot Interaction with Dialogs Containing Irony, Sarcasm and Jocularity

The goal-directed behavior observed in humans arises from the intricate interplay of various processes, including personality dynamics, emotional responses to others, memory encoding, the anticipation of future actions, and associated hedonic experiences. Integrating these multiple processes characteristic of human intelligence into a robotic framework aims to enhance the human-likeness of artificial agents and facilitate more natural and intuitive interactions with humans.   For this purpose, in this paper, we propose a comprehensive psychological and cognitive architecture where, personality, as it happens for humans, not only influences the execution of actions but also shapes internal reactions to human emotions and guides anticipatory decision-making processes tailored to the agent's traits. We demonstrate the framework's effectiveness in generating perceivable synthetic personalities through an experiment involving participants in a dyadic conversation scenario with a digital human, where the digital human's behavior is driven by its assigned personality. The results show that participants accurately perceive the artificial personality displayed by the digital human. We also demonstrate the potential of our robotic framework to bridge the gap between cognitive and psychological agents, as the findings highlight its ability to create a cognitively and emotionally intelligent digital human.

Personality and Memory-Based Framework for Emotionally Intelligent Agents

To enable sophisticated interactions between humans and robots in a shared environment, robots must infer the intentions and strategies of their human counterparts. This inference can provide a competitive edge to the robot or enhance human-robot collaboration by reducing the necessity for explicit communication about task decisions. In this work, we identify specific states within the shared environment, which we refer to as Critical Decision Points, where the actions of a human would be especially indicative of their high-level strategy. A robot can significantly reduce uncertainty regarding the human's strategy by observing actions at these points. To demonstrate the practical value of Critical Decision Points, we propose a Receding Horizon Planning (RHP) approach for the robot to influence the movement of a human opponent in a competitive game of hide-and-seek in a partially observable setting. The human plays as the hider and the robot plays as the seeker. We show that the seeker can influence the hider to move towards Critical Decision Points, and this can facilitate a more accurate estimation of the hider's strategy. In turn, this helps the seeker catch the hider faster than estimating the hider's strategy whenever the hider is visible or when the seeker only optimizes for minimizing its distance to the hider.

Planning with Critical Decision Points: Robots That Influence Humans to Infer Their Strategy

Despite the rise of mobile robot deployments in home and work settings, perceived safety of users and bystanders is understudied in the human-robot interaction (HRI) literature. To address this, we present a study designed to identify elements of a human-robot encounter that correlate with observed stress response. Stress is a key component of perceived safety and is strongly associated with human physiological response. In this study a Boston Dynamics Spot and a Unitree Go1 navigate autonomously through a shared environment occupied by human participants wearing multimodal physiological sensors to track their electrocardiography (ECG) and electrodermal activity (EDA). The encounters are varied through several trials and participants self-rate their stress levels after each encounter. The study resulted in a multidimensional dataset archiving various objective and subjective aspects of a human-robot encounter, containing insights for understanding perceived safety in such encounters. To this end, acute stress responses were decoded from the human participants' ECG and EDA and compared across different human-robot encounter conditions. Statistical analysis of data indicate that on average (1) participants feel more stress during encounters compared to baselines, (2) participants feel more stress encountering multiple robots compared to a single robot and (3) participants stress increases during navigation behavior compared with search behavior.

Human Stress Response and Perceived Safety During Encounters with Quadruped Robots

This study employs both qualitative and quantitative approaches to explore perceptions of sex robots. Utilizing grounded theory, a thematic analysis revealed three primary themes: uncanny, artificial, and curious. Following this, an exploratory factor analysis was conducted to evaluate the dimensionality of these themes. Regression analyses were then performed to examine the influence of individual differences on these dimensions. The findings indicated that females perceived sexbots as more uncanny, older individuals perceived sexbots as more artificial, Americans were more curious about sexbots than the British, and males were more curious about sexbots, particularly the female sexbots.

Uncanny, Artificial, but Curious: A Mixed Methods Study on Responses to Sex Robots

Understanding prosociality toward robots furthers our insight into robot acceptance and perception. This study (N=310) adopts a third-person perspective to investigate outcomes of the Dictator Game paradigm. Participants were tasked with evaluating the shared allocation of four distinct resources (money, electronic tools, food, and access to electricity). Results reveal that only 27% of answers indicate a preference for non-sharing. Our findings also show a tendency towards greater sharing of electronic tools and lesser sharing of food compared to access to electricity. Notably, perceptions of robot trustworthiness do not influence peoples sharing judgments, while anthropomorphic perception does play a significant role. Participants expressed greater confidence in their judgments regarding sharing food with robots, a resource where participants preferred non-sharing. This suggests that individuals most likely distinctly categorize robots when contemplating resources necessary for sustaining life. Our study indicates an emerging societal expectation regarding the types and extent of resources to be shared with robots, thus contributing to the delineation of evolving robot social norms.

Judging Sharing with Robots: A Third-Person Perspective on Prosociality

This study investigates the impact of the Internet of Robotic Things (IoRT) on user perceptions when the system encourages users behavior in accordance with social norms. Utilizing a 2 (Relationship between humans: hierarchical vs. flat) x2 (IoRT mode: hierarchical vs. flat) mixed-participants design, the research explores how different modes of IoRT operationflat and hierarchicalaffect perceived usefulness, service evaluation of the IoRT, and perceived sociability and familiarity towards counterparts. Results indicated that the flat mode was rated more favorably than the hierarchical mode in terms of usefulness and service evaluation. Additionally, sociability and familiarity ratings towards the interaction counterpart significantly varied based on the relationship type and IoRT mode, with the flat mode enhancing these perceptions particularly in flat relationship conditions.

The Effect of IoRT on Technology Perception and User Impressions through Social Norm Compliance

Endoscopic retrograde cholangiopancreatography (ERCP) is a minimally invasive technique for treating biliary and pancreatic diseases. It requires inserting a guide wire into the bile duct or pancreatic duct to enable further procedures. However, this is challenging because the duct openings are at an obtuse angle to the direction of the instrument entering the duodenum. The guide wire may fail to adjust its position and posture, causing mis-insertion and tissue damage. To address this issue, we proposed a flexible surgical robot with a concentric design that combines magnetic-driven and tendon-driven methods. The NiTi wires work as both the backbone and the driving element of the outer tube, and a flexible inner part equipped with several magnets works as a guide wire that can bend in random directions. By using two steerable segments, our robot has higher flexibility and a lower chance of tissue damage. Then, we analyse the dexterity of the proposed mechanism and performed a phantom experiment. The result shows that it can perform insertion with different postures and adapt to the natural configuration of the natural cavity of a human.

Magnetic Concentric Guidewire Inserting Robot towards Endoscopic Retrograde Cholangiopancreatography

This paper introduces an enhanced hand gesture recognition system utilizing inductive sensors, specifically developed for the identification of American Sign Language (ASL) gestures. The system is built into a regular glove with conductive threads by sewing coils onto various areas, including the fingers, wrist, palm, and ulnar regions. To increase the system's sensitivity, a data acquisition system measures a tank circuit created by each sewn coil and external components. The effectiveness of the system is rigorously evaluated on a sample of 10 participants. The identification of the 26 ASL letters is achieved using a random forest machine learning algorithm. To enhance the diversity and applicability of the machine learning model, a data augmentation scheme is utilized. The outcomes reveal a notable level of accuracy, achieving 98.91% accuracy with 5-fold cross-validation and 98.82% accuracy with leave-one-subject-out cross-validation.

Premium content

Downloads

Next from IEEE RO-MAN 2024

A Biologically Inspired Program-Level Imitation Approach for Robots: Proof-Of-Concept

Similar lecture

Learning Diffusion Policy from Primitive Skills for Robot Manipulation