Overview

Messages from electronic devices are conventionally received as text, audio, or radio signals. But robots move with rich, articulate motion in the real world, opening up the possibility of transmitting messages through motion itself.

We introduce messaging through motion: a method for modifying a robot's movements to transmit arbitrary messages — such as an agent's current intent — as noise in any pre-trained policy's actions, detectable from remote sensing (e.g. video or motion capture) without degrading policy performance. The channel complements existing wireless comms but does not depend on them, requiring no extra hardware or direct link to the robot.

On real robots running at 50 Hz, we achieve 100% message recovery at 0.67 bits/s.

Design Space

• Alphabet size M: (larger = more bits/symbol)

• Time per symbol T_sym: (longer = more accurate)

• Channel strategy: 

  • • Single channel (simple baseline)

    • Same-message (send the same message across multiple channels)

    • Split-message (send different parts of the same message across mutliple channels)

Simulation Experiments 

We validate our approach across four environments of increasing difficulty, ranging from simple navigation to complex multi-agent adversarial games.

Real-world Experiments 

Results

We evaluate decoding accuracy across all platforms. As the symbol duration increases, the frequency-domain watermark becomes increasingly resilient to noise.