Discuss HRI

[LeroyR]

The current HRI tasks at RoboCup are somewhat limited and do not fully explore interesting or impactful research topics. Let’s discuss what could make a new HRI task both challenging and meaningful for teams and for research.

---

Observations from previous competitions:

• Robots often struggle with interaction and are not sufficiently “attentive” to users.
• Referees currently cannot ask the robot to repeat or rephrase instructions at any time.
• During tasks, we observed timeouts while volunteers are being instructed, why can’t the robot detect uncertainty or allow the referee to ask it to wait?
• Robots rarely detect social cues or uncertainty from participants, leading to incomplete or awkward interactions.

Goals for a potential new task:

• Focus on interaction-heavy scenarios, beyond simple token dialogues (e.g., “What is your name?”).
• Incorporate multiple modalities, including: Speech Gestures (wave, raise hand), Gaze and pointing, Social cues
• Be measurable and competitive, so performance can be objectively evaluated.

After quick scan, this could be research directions:

• Multi-user dialog management
• Full-duplex systems for natural conversation
• Attention and social cue recognition
• Adaptive interaction strategies based on user uncertainty

[LeroyR]

Older Ideas: #941

[moriarty]

Robots often struggle with interaction and are not sufficiently “attentive” to users.

Robots were also seen driving away from user while attempting to communicate with them, giving no social cue that to expect dialogue.