Lifelong Learning and Personalization in Long-Term Human-Robot Interaction (LEAP-HRI)

About
Global inequalities in access to essential resources such as education, healthcare, and technology continue to widen social and economic disparities, especially in underserved and underrepresented communities. The growing integration of foundation models and other machine learning systems in robots offers promising and personalized solutions that can adapt to various individuals, situations, and environments, potentially addressing some of these gaps. By learning from interactions and evolving with local conditions, these systems can provide individualized support, such as assisting older adults with daily tasks, aiding children with special needs in learning environments, or empowering people with disabilities to live more independently. Building trust and fostering collaboration between humans and robots will help ensure that these systems meet the unique needs of all individuals, especially within long-term humanrobot interaction (HRI).

With this year’s theme of “Overcoming Inequalities with Adaptation”, in line with the overall theme of the conference “Robots for a Sustainable World”, the fifth edition of the “Lifelong Learning and Personalization in Long-Term Human-Robot Interaction (LEAP-HRI)” workshop aims to bring together insights across diverse disciplines, exploring how continually evolving robots can effectively operate in diverse environments, promoting greater equity, inclusivity, and empowerment for individuals and communities. The workshop aims to facilitate collaborations across diverse scientific perspectives through a keynote presentation, panel discussions, and in-depth discussions on the contributed talks, attempting to shape a more sustainable and equitable future through adaptive advancements in long-term HRI.

Schedule

Time (GMT-6)
09:00 am - 09:10 am	Introductory Remarks: Bahar Irfan
09:10 am - 09:40 am	Keynote by Dana Kulic: Human Variability and Long-term HRI Abstract While there are many challenges to long term human-robot interaction, one of the key challenges is how to handle human variability over time. Human variability can arise from both the need for the robot to interact with diverse users, and from changes to a single user's profile over time, e.g., due to ageing, skill improvement, or habituation. This talk will describe two types of approaches to handling variability: designing policies that adapt to individual users, and designing policies that are robust to variations in users. Example HRI applications and domains will be highlighted, including warehouse robot supervision, gait rehabilitation and shared autonomy drone flight.	Chair: Nikhil Churamani
09:40 am - 09:50 am	Long-Term Planning Around Humans in Domestic Environments [PDF] Authors: Ermanno Bartoli, Dennis Rotondi, Iolanda Leite Abstract Long-term planning for robots operating in domestic environments poses unique challenges due to the interactions between humans, objects, and spaces. Recent advancements in trajectory planning have leveraged vision-language models (VLMs) to extract contextual information for robots operating in real-world environments. While these methods achieve satisfying performance, they do not explicitly model human activities. Such activities influence surrounding objects and reshape spatial constraints. This paper presents a novel approach to trajectory planning that integrates human preferences, activities, and spatial context through an enriched 3D scene graph (3DSG) representation. By incorporating activity-based relationships, our method captures the spatial impact of human actions, leading to more context-sensitive trajectory adaptation. Preliminary results demonstrate that our approach effectively assigns costs to spaces influenced by human activities, ensuring that the robot’s trajectory remains contextually appropriate and sensitive to the ongoing environment. This balance between task efficiency and social appropriateness enhances context-aware human-robot interactions in domestic settings. Future work includes implementing a full planning pipeline and conducting user studies to evaluate trajectory acceptability.	Chair: Michelle Zhao
09:50 am - 10:00 am	A Data-Driven Framework for Skill Representation [PDF] Authors: Mariah Schrum, Allison Morgan, Deepak Gopinath, Jean Costa, Emily Sumner, Guy Rosman, Tiffany Chen Abstract Understanding and modeling human skill is critical for advancing human-robot and human-AI interaction, particularly in domains requiring nuanced cooperation and long-term personalization. Effective collaboration depends on aligning AI behavior with human capabilities, but quantifying skill is challenging, requiring both task knowledge and expert intuition. This work proposes a data-driven approach to modeling human skill through repeated interactions. Using high-performance driving education (HPDE) as a case study, we synthesize literature and expert insights to identify supervision signals for learning a robust skill representation. Leveraging a dataset of novice and expert drivers, we demonstrate the feasibility of automatically extracting skill representations from track-driving trajectory data. This foundation enables applications such as personalized coaching, skill-targeted challenges, and adaptive robotic interventions. Ultimately, we aim to democratize HPDE by making high-quality instruction more accessible through AI-driven personalization.
10:00 am - 10:10 am	Task-Relevant Active Learning without Prior Knowledge using Vision-Language Models [PDF] Authors: Usman Irshad Bhatti, Ali Ayub Abstract Traditional active learning methods rely on uncertainty-based selection, which assumes the presence of initial labeled data to guide sample selection. These methods also assume that all the unknown data is relevant to the task to be performed by the robot, which is untrue in real world settings. This problem becomes more challenging when robots only know the names of relevant classes for their tasks but have no labeled data, rendering uncertainty sampling and other informative selection strategies infeasible. The only practical method in these situations is random sampling, which results in an ineffective use of labeling resources. As a solution to this problem, we suggest a task-relevant active learning method in which we pre-select samples that are most probably from relevant classes by using embeddings generated by vision-language models and calculating similarity scores between written task descriptions and object images. This minimizes annotation waste by ensuring that the labeled dataset begins with a high percentage of pertinent samples. According to experimental findings, our approach outperforms random sampling and traditional active learning in terms of model performance and relevant data selection, which makes it a workable option for robotic learning systems that have no initial labels and limited data to learn from in the real world.
10:10 am - 10:30 am	Break-out Session #1	Chair: Ali Ayub
10:30 am - 11:00 am	Coffee break
11:00 am - 11:50 am	Panel: Overcoming Inequalities with Adaptation	Moderator: Silvia Rossi Online Chair: Michelle Zhao
11:50 am - 12:15 am	Break-out Session #2	Chair: Ali Ayub
12:15 am - 12:30 am	Break-out Wrap-up and Concluding Remarks	Chair: Bahar Irfan

Call for Papers
The call for papers is now closed.

Areas of interest

We encourage researchers and students from HRI, robotics, cognitive science, rehabilitation, social sciences, and educational backgrounds to contribute. The workshop welcomes contributions across a wide range of topics including, but not limited to:

Lifelong personalization and/or adaptation
Bias mitigation in adaptive HRI
Privacy and ethical considerations in lifelong learning/ personalization in HRI
Cross-cultural adaption in HRI
Achieving fairness and inclusivity with learning and adaptation
Incremental and/or online learning in HRI
Modeling user(s) and/or user behavior(s) in multi-session/long-term HRI
Modeling robot behavior in multi-session/long-term HRI
Modeling context in multi-session/long-term HRI
Agent/robot architectures for personalization/adaptation
Lifelong (long-term) human-agent or multi-user/multi-agent interactions
Lifelong (long-term) multimodal interactions
Continual/lifelong machine learning

Submission Guidelines

We invite scientific papers ranging from 3 to 4 pages, with additional space allocated for references and appendices. Submissions can encompass various types of work, including ongoing projects with preliminary findings, technical reports, case studies, surveys, and cutting-edge research in the realms of lifelong learning and personalization. These topics span diverse fields in real-world applications, such as education, rehabilitation, elderly care, collaborative tasks, customer-oriented services, and companion robots, as well as long-term studies.

We encourage submissions to align their submissions with the overarching theme of the workshop, “Overcoming Inequalities with Adaptation”. All submitted papers will undergo a thorough review process to assess their relevance, originality, and scientific and technical robustness.

Submissions do not need to be anonymized for review. All manuscripts must be written in English and submitted electronically in PDF format via EasyChair. The accepted papers will be published on the workshop website, as well as in arXiv. Note that at least one author needs to register to HRI with (or only) workshop option, and attend in person or online for the accepted paper to be presented at the workshop.

In line with HRI full conference paper formatting, authors should use IEEE Conference format: Template files (US letter) or Overleaf template.

Important Dates

Early-bird submission deadline: January 24
Early-bird notification of acceptance: January 30
General submission deadline: February 14
General notification of acceptance: February 21
Camera-ready deadline: February 26

All deadlines are at 23:59 Anywhere on Earth time. Deadlines will not be extended.