Interdisciplinary field studying how humans and robots communicate, collaborate, and interact safely and effectively in shared physical and virtual spaces. HRI combines robotics engineering, AI, human factors engineering, and cognitive science to design systems with natural interaction modalities including gesture recognition, natural language processing, and haptic feedback.

Semantic Classification

Content

Current Landscape (2026)

  • Vision-language-action (VLA) foundation models now drive natural-language HRI: Google DeepMind released Gemini Robotics and Gemini Robotics-ER (March 2025, built on Gemini 2.0) and followed with Gemini Robotics 2, while Physical Intelligence’s flow-matching pi-0 (Oct 2024) and hierarchical pi-0.5 (2025), and NVIDIA’s dual-system GR00T N1 (2025), let non-experts instruct robots conversationally rather than scripting them.
  • General-purpose humanoids moved from demos into workplace pilots: Figure AI’s Figure 03 (with an OpenAI-partnered reasoning stack and its Helix/Helix-02 VLA, which showed the first multi-robot collaborative loco-manipulation), 1X’s Neo learning from human demonstration, and Boston Dynamics’ Electric Atlas using zero-shot sim-to-real whole-body skills.
  • Deployment milestones include Toyota contracting seven Agility Digit units for a Canadian factory (Feb 2026) and Tesla exhibiting Optimus v2.3 publicly outside the US for the first time (London 13 Dec 2025, Berlin 20 Dec 2025), greeting and handing items to attendees.
  • Safety standardisation was overhauled: ISO 10218-1:2025 and ISO 10218-2:2025 came into force on 1 April 2025, absorbing the former ISO/TS 15066 collaborative contact-force limits directly into the integrator standard and adding cybersecurity requirements; the US adopted them as ANSI/A3 R15.06-2025 (published Sept-Oct 2025).
  • A dedicated humanoid safety standard, ISO 25785-1 for dynamically stable (walking) robots, remained a Working Draft as of January 2026 (working group led by Agility Robotics, Boston Dynamics and the A3 Association, last met in Barcelona Oct 2025), with publication expected 2026-2027 and covering fall-zone calculations.
  • Regulation is tightening around interactive and AI-driven robots: the EU Machinery Regulation 2023/1230 applies from 20 January 2027 (adding autonomous-machinery and collaborative-robot provisions), and the EU AI Act (Regulation 2024/1689) can classify a robot’s safety-critical AI functions as high-risk.
  • Investment surged, with humanoid-focused funding exceeding an estimated 1.3 billion US dollars in H1 2025 (Figure AI’s 1 billion dollar round, Apptronik ~403 million, Agility ~400 million), even as Tesla’s Optimus programme faced a mid-2025 leadership departure and a trade-secret lawsuit and slipped its promised 2026 sales start.
  • Open challenges as of 2026 remain reliable autonomy without teleoperation, safe close-contact collaboration and fall risk for legged humanoids, trust and legibility of robot intent, and certifying learned VLA policies whose behaviour is not exhaustively specified.

References

Provenance