Detecting and responding to sirens

A vehicle configured to operate in an autonomous driving mode, the vehicle comprising:

• a perception system including plurality of synchronized microphones configured to detect a noise emanating from another vehicle;
• and
• one or more processors operably coupled with the perception system, the one or more processors configured to:
• form, by beamforming an array of at least a subset of the plurality of synchronized microphones, a plurality of beams in different directions external to the vehicle;
• and
• apply sound corresponding to each of the formed plurality of beams to a first model, wherein the first model focuses listening on one or more locations that are most relevant to the other vehicle.

A method for controlling a vehicle in an autonomous driving mode, the method comprising:

• using, by one or more processors, output from a perception system including plurality of synchronized microphones configured to detect a noise emanating from another vehicle to generate probabilities of the noise emanating from a particular direction relative to the vehicle configured to operate in an autonomous driving mode;
• estimating, by the one or more processors, a bearing of the other vehicle based on the generated probabilities;
• identifying, by the one or more processors, at least one road segment associated with the other vehicle;
• and
• controlling, by the one or more processors, the vehicle in the autonomous driving mode to perform a maneuver in response to the estimated bearing and the identified at least one road segment.