Interactive and Explainable Robot Learning: A Comprehensive Review

Esmaeil Seraj, Kin Man Lee, Zulfiqar Zaidi, Qingyu Xiao, Zhaoxin Li, Arthur Nascimento, Sanne van Waveren, Pradyumna Tambwekar, Rohan R. Paleja, Devleena Das, Matthew C. Gombolay

Abstract: This review embarks on a comprehensive exploration of approaches, evaluation methods, and ethical considerations in explainable and interactive systems for robotic applications, distinctly focusing on intelligent systems that are specifically designed for learning automated agents. Given the increasing integration of robots in daily life, it is crucial to focus on intelligent systems that can not only learn and adapt, but can also offer clarity and comprehension for their actions. The interactive component of these systems is thoroughly examined, evaluating the algorithms, the modalities used in interaction, and the significance of mixed-initiative and shared autonomy. We spotlight adaptive and adaptable methods, emphasizing the centrality of user-inspired research and personalized approaches in interactive robotics.A rigorous examination of safety and ethical considerations of these intelligent systems anchors the discussion, including aspects of transparency, privacy, accountability, biases, and psychological well-being. The review evaluates existing metrics and benchmarking standards for such systems and explores their practical applications across domains such as healthcare, domestic tasks, and industrial automation. Concluding with key insights and directions for future research, we provide design guidelines and points of consensus for each subject in order to equip readers with a nuanced understanding of current trends and tools in explainable and interactive robotic systems, paving the way for informed research and application in this dynamic field.

External IDs:dblp:journals/ftrob/SerajLZXLNWTPDG24