Robot learning from humans has been proposed and researched for several decades as a means to enable robots to learn new skills or adapt existing ones to new situations. Recent advances in AI, including learning approaches like reinforcement learning and architectures like transformers and foundation models, combined with access to massive datasets, have created attractive opportunities to apply those data-hungry techniques to this problem. We argue that the focus on massive amounts of pre-collected data, and the resulting learning paradigm, where humans demonstrate and robots learn in isolation, is overshadowing a specialized area of work we term Human-Interactive Robot Learning (HIRL). This paradigm, wherein robots and humans interact during the learning process , is at the intersection of multiple fields (AI, robotics, human–computer interaction, design and others) and holds unique promise. Using HIRL, robots can achieve greater sample efficiency (as humans can provide task knowledge through interaction), align with human preferences (as humans can guide the robot behavior toward their expectations), and explore more meaningfully and safely (as humans can utilize domain knowledge to guide learning and prevent catastrophic failures). This can result in robotic systems that can more quickly and easily adapt to new tasks in human environments. The objective of this article is to provide a broad and consistent overview of HIRL research and to guide researchers toward understanding the scope of HIRL, and current open or underexplored challenges related to four themes—namely, human, robot learning, interaction, and broader context. The article includes concrete use cases to illustrate the interaction between these challenges and inspire further research according to broad recommendations and a call for action for the growing HIRL community.

Training a robot through demonstration requires robust algorithms capable of processing provided trajectories to generate high-quality task executions. However, the success of this process is highly dependent on the quality of the trajectories. Poor-quality trajectories hinder the ability of algorithms to learn and generalize effectively, while high-quality trajectories can improve learning outcomes, reducing the need for overly complex algorithms. In this work, we propose an initial method for comparing sets of semantically annotated trajectories used for robot demonstration. To evaluate the proposed methodologies, we recorded trajectories of 60 participants in a study setup in Virtual Reality where each participant tested one of three different control-visualization compositions. We use classical approaches such as Dynamic Time Warping and Discrete Frechet-Distance to measure the similarity between segments of recorded trajectories and show that different input devices and visualization combinations affect the resulting trajectory metrics.

AURORAis a software platform, that facilitates scalable deployment of robotic simulations over the web for the Human-Robot Interaction (HRI) community. As robotics is becoming increasingly important in various disciplines, there is a growing need for accessible and scalable research methods. Traditional experiments often require expensive hardware and in-person participation, limiting accessibility and participant diversity. Our platform allows researchers from different fields to easily provide HRI experiences by deploying online studies with robotic simulations paired with customizable surveys, allowing end users worldwide to interact with these simulations. Our platform is entirely open source and can be hosted locally, providing flexibility and control of the research environment. Since AURORA is implemented with Docker, it is platform-independent. By offering a user-friendly interface that can be deployed and used without extensive technical expertise, our platform reduces costs, increases participant diversity, and improves the reproducibility of research in the HRI community.

The ability to generate explanations that are understood by explainees is the quintessence of explainable artificial intelligence. Since understanding depends on the explainee’s background and needs, recent research focused on co-constructive explanation dialogues, where an explainer continuously monitors the explainee’s understanding and adapts their explanations dynamically. We investigate the ability of large language models (LLMs) to engage as explainers in co-constructive explanation dialogues. In particular, we present a user study in which explainees interact with an LLM in two settings, one of which involves the LLM being instructed to explain a topic co-constructively. We evaluate the explainees’ understanding before and after the dialogue, as well as their perception of the LLMs’ co-constructive behavior. Our results suggest that LLMs show some co-constructive behaviors, such as asking verification questions, that foster the explainees’ engagement and can improve understanding of a topic. However, their ability to effectively monitor the current understanding and scaffold the explanations accordingly remains limited.

This paper introduces the Mental Model Mismatch (MMM) Score, a feedback mechanism designed to align human teaching behavior with robot learning by quantifying mismatches between the human teacher’s mental model and the robot’s learning capabilities. Utilizing a LLM, the system analyzes human teacher intentions in natural language to generate adaptive feedback for the human. A study with 150 participants teaching a virtual robot learner demonstrates that intention-based feedback significantly improves the robots learning outcomes compared to traditional performance-based feedback or no feedback. The findings suggest that this approach improves understanding of the robot’s learning process and reduces misconceptions to enhance human-robot interaction.

This work examines the critical role of nonverbal cues in initiating successful human-robot interactions (HRI), focusing on the development of an algorithm capable of decoding subtle cues that indicate a human’s intention to interact with a robot. The foundation for the analysis is the concept of closeness which relates to the individual feeling of intimacy and comfort, and indicates the engagement potential in diverse settings and interactions. This paper introduces a novel algorithm designed to quantify a person’s willingness to interact based on analyzing the displayed closeness on a series of images. The algorithm’s efficacy is evaluated through an empirical study involving 20 participants, generating a substantial dataset of 130 video sequences. These sequences were analyzed to assess the algorithm’s ability to predict communication initiation intentions, with findings suggesting a significant distinction between participants engaged in interaction with a robot and those instructed to ignore it. The results underscore the algorithm’s potential in facilitating more nuanced and socially sustainable HRI, laying the groundwork for future advancements in the field. This research contributes to the growing body of work on social robotics, emphasizing the importance of integrating nonverbal cue analysis for enhancing robots’ interactive capabilities and fostering more meaningful human-robot connections.

In this contribution, we describe a method of analysing and interpreting the direction and timing of a human's gaze over time towards a robot whilst interacting. Based on annotated video recordings of the interactions, this post-hoc analysis can be used to determine how this gaze behaviour changes over the course of an interaction, following from the observation that humans change their behaviour towards the robot on the time-scale of individual interactions. We posit that given these circumstances, this measure may be used as a proxy (among others) for engagement in the interaction or the human's attribution of social agency to the robot. Application of this method to a sample of unstructured child-robot interactions demonstrates its use, and justifies its utilisation in future studies.

Robots are not human. They might in some cases have a similar appearance but different behavioral and cognitive strengths and limitations. In this sense, an interaction with a robot is asymmetric. When interacting with a robot one is unsure what behavior to expect as the appearance does not necessarily make the abilities of the robot transparent. In human-human interaction, we can also find asymmetric interactions to occur. For example, in an interaction with a child, adults have to adapt to the learner's capabilities and understanding. Similarly, in interactions with special populations such as persons with autistic spectrum disorders (ASD), asymmetry occurs as specific information seems to be processed differently.

This paper presents a new method for detecting and merging redundant points in registered range data. Given a global representation from sequences of 3D points, the points are projected onto a virtual image plane computed from the intrinsic parameters of the sensor. Candidates for redundancy are collected per pixel which then are clustered locally via region growing and replaced by the clusterpsilas mean value. As data is provided in a certain manner defined by camera characteristics, this processing step preserves the structural information of the data. For evaluation, our approach is compared to two other algorithms. Applied to two different sequences, it is shown that the presented method gives smooth results within planar regions of the point clouds by successfully reducing noise and redundancy and thus improves registered range data.