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.

Human-in-the-loop learning is gaining popularity, particularly in the field of robotics, because it leverages human knowledge about real-world tasks to facilitate agent learning. When people instruct robots, they naturally adapt their teaching behavior in response to changes in robot performance. While current research predominantly focuses on integrating human teaching dynamics from an algorithmic perspective, understanding these dynamics from a human-centered standpoint is an under-explored, yet fundamental problem. Addressing this issue will enhance both robot learning and user experience. Therefore, this paper explores one potential factor contributing to the dynamic nature of human teaching: robot errors. We conducted a user study to investigate how the presence and severity of robot errors affect three dimensions of human teaching dynamics: feedback granularity, feedback richness, and teaching time, in both forced-choice and open-ended teaching contexts. The results show that people tend to spend more time teaching robots with errors, provide more detailed feedback over specific segments of a robot’s trajectory, and that robot error can influence a teacher’s choice of feedback modality. Our findings offer valuable insights for designing effective interfaces for interactive learning and optimizing algorithms to better understand human intentions.

Reinforcement Learning (RL) is an effective method for robots to learn tasks. However, in typical RL, end-users have little to no control over how the robot does the task after the robot has been deployed. To address this, we introduce the idea of online behavior modification, a paradigm in which users have control over behavior features of a robot in real-time as it autonomously completes a task using an RL-trained policy. To show the value of this user-centered formulation for human-robot interaction, we present a behavior-diversity--based algorithm, Adjustable Control Of RL Dynamics (ACORD), and demonstrate its applicability to online behavior modification in simulation and a user study. In the study (n =23), users adjust the style of paintings as a robot traces a shape autonomously. We compare \algoshort to RL and Shared Autonomy (SA), and show \algoshort affords user-preferred levels of control and expression, comparable to SA, but with the potential for autonomous execution and robustness of RL. The code for this paper is available at https://github.com/AABL-Lab/HRI2024_ACORD

It is crucial that users are empowered to take advantage of the functionality of a robot and use their understanding of that functionality to perform novel and creative tasks. Given a robot trained with Reinforcement Learning (RL), a user may wish to leverage that autonomy along with their familiarity of how they expect the robot to behave to collaborate with the robot. One technique is for the user to take control of some of the robot’s action space through teleoperation, allowing the RL policy to simultaneously control the rest. We formalize this type of shared control as Partitioned Control (PC). However, this may not be possible using an out-of-the-box RL policy. For example, a user’s control may bring the robot into a failure state from the policy’s perspective, causing it to act unexpectedly and hindering the success of the user’s desired task. In this work, we formalize this problem and present Imaginary Out-of-Distribution Actions, IODA, an initial algorithm which empowers users to leverage their expectations of a robot’s behavior to accomplish new tasks. We deploy IODA in a user study with a real robot and find that IODA leads to both better task performance and a higher degree of alignment between robot behavior and user expectation. We also show that in PC, there is a strong and significant correlation between task performance and the robot’s ability to meet user expectations, highlighting the need for approaches like IODA. Code is available at https://github.com/AABL-Lab/ioda_roman_2024

Interactive Reinforcement Learning (IntRL) allows human teachers to accelerate the learning process of Reinforcement Learning (RL) robots. However, IntRL has largely been limited to tasks with discrete-action spaces in which actions are relatively slow. This limits IntRL's application to more complicated and challenging robotic tasks, the very tasks that modern RL is particularly well-suited for. We seek to bridge this gap by presenting Continuous Action-space Interactive Reinforcement learning (CAIR): the first continuous action-space IntRL algorithm that is capable of using teacher feedback to out-perform state-of-the-art RL algorithms in those tasks. CAIR combines policies learned from the environment and the teacher into a single policy that proportionally weights the two policies based on their agreement. This allows a CAIR agent to learn a relatively stable policy despite potentially noisy or coarse teacher feedback. We validate our approach in two simulated robotics tasks with easy-to-design and - understand heuristic oracle teachers. Furthermore, we validate our approach in a human subjects study through Amazon Mechanical Turk and show CAIR out-performs the prior state-of-the-art in Interactive RL.