Reproduce “Interactive Perception: Leveraging Action in Perception and Perception in Action” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Jeannette Bohg, Karol Hausman, Bharath Sankaran, Oliver Brock, Danica Kragic, Stefan Schaal, Gaurav Sukhatme

This survey paper consolidates the field of Interactive Perception (IP) in robotics, proposing that robots can enhance perception and manipulation by actively interacting forcefully with their environment. It highlights how these interactions generate novel sensory data and reveal predictable action-perception regularities, simplifying complex tasks.

Key insights

  • Embodied robots can transform passive observation into active exploration by generating informative sensory data through purposeful physical engagement.
  • The predictability of sensory changes resulting from specific actions offers a powerful mechanism to simplify perception and reduce data requirements compared to static analysis.
  • IP unifies and extends prior concepts like Active Perception and Active Haptic Perception by emphasizing deliberate, forceful interaction for richer information gain.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.