A historical review arguing that a complete artificial agent must include active perception: choosing how to move and where to look in order to perceive better. The authors trace the ideas back through decades of robotics and vision work, suggest that limited compute and costly hardware held them back, and argue they are worth reviving with modern tools.
Findings
- A review with no experiments. The authors admit their history may be selective.
Limits
- Abstract-level reading only.
Relevance
Moving to reduce uncertainty is what GoGet does when it approaches an object after its models disagree. The multi-robot version is open: when two robots disagree about an object, which one should move, and is it cheaper to move or to ask a third? That question feeds multi-robot multimodal perception and long-term autonomy; see also query by committee.
Links
- arXiv
- Subfield: Perception, Planning and navigation