A Scoping Review and Experimental Study on Reinforcement Learning from Human Feedback for Human-Robot Collaboration
Review of 20 RLHF human-robot papers finds user-initiated feedback better captures perceived navigation safety.
A PRISMA scoping review screened 199 records and included 20 peer-reviewed papers from 2020 to 2025 on reinforcement learning from human feedback for human-robot collaboration. It maps feedback modalities, where collected data enters training, and pilot evaluations that use both human and robot metrics. A between-subjects VR experiment compared system-initiated and user-initiated feedback on robot proxemic behavior for safe navigation. Bayesian analysis found user-initiated feedback better captured psychological safety, while physical safety was measured by inverse time-to-collision.
- PRISMA review included 20 of 199 records from 2020-2025
- Feedback can enter training at multiple stages and formats
- User-initiated feedback matched perceived safety better than system prompts
- Physical safety was measured with inverse time-to-collision
From the original article
1 min read at arxiv.orgHuman-Robot Collaboration (HRC) can facilitate mass customisation in Industry 4. 0, with Reinforcement Learning from Human Feedback (RLHF) representing a promising approach for developing safe AI-based robots. Practical challenges remain regarding safety during AI development, human feedback quality, and bidirectional human-robot adaptation.
By Alexandra Coroiu · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.