Robot attention detection

1. A robot comprising:

a body and one or more physically moveable components;

one or more sensor subsystems that are configured to generate sensor inputs comprising respective representations of an environment of the robot;

one or more processors; and

one or more storage devices storing instructions that are operable, when executed by the one or more processors, to cause the robot to perform operations comprising:

computing a plurality of attention signals from the sensor inputs, wherein each attention signal corresponds to one factor of multiple factors for classifying an attention level using an attention level classifier of the robot according to the sensor inputs,

providing the plurality of attention signals as input to the attention level classifier to generate an attention level

determining that a user is paying attention to the robot based on the generated attention level;

in response, determining a behavior to execute based on the current attention level, wherein a behavior comprises one or more coordinated actions to be performed by the robot, and

executing the determined behavior including performing the one or more coordinated actions.

2. The robot of claim 1 , wherein executing the determined behavior comprises uploading a stream of received audio for natural language processing without having received a wake word.

3. The robot of claim 1 , wherein the attention level represents a predicted likelihood that a user is paying attention to the robot based on the sensor inputs.

4. The robot of claim 1 , wherein the behavior causes the robot present an eye animation on a display to acknowledge the user.

5. The robot of claim 1 , wherein determining a behavior to execute based on the current attention level comprises determining a physical behavior to execute, and wherein executing the determined physical behavior effects a physical movement of one or more of the physically moveable components of the robot.

6. The robot of claim 5 , wherein the one or more sensor subsystems comprise an audio subsystem configured to capture audio from the environment of the robot and to determine a direction from which the audio originated,

and wherein the determined physical behavior comprises one or more actions directing the robot to turn toward the direction from which the audio originated.

7. The robot of claim 5 , wherein the one or more sensor subsystems comprise an audio subsystem configured to capture audio from the environment of the robot and to determine a direction from which the audio originated,

and wherein the determined physical behavior comprises one or more actions directing the robot to drive forward in the direction from which the audio originated.

8. The robot of claim 1 , wherein the one or more sensor subsystems comprise a perception subsystem configured to sense light from an environment of the robot, and

wherein the attention signals comprise:

a first indication of whether a human face is recognized in image data captured by the perception subsystem,

a second indication of whether a gaze is detected for at least a threshold amount of time in image data captured by the perception subsystem, or

a third indication of whether mouth movement is detected in image data captured by the perception subsystem.

9. The robot of claim 8 , wherein the first indication is an indication of whether a human face of a known user is recognized.

10. The robot of claim 1 , wherein the one or more sensor subsystems comprise a perception subsystem configured to sense light from an environment of the robot and an audio subsystem configured to capture audio from the environment of the robot and to determine a direction from which the audio originated, and

wherein the attention signals comprise:

a first indication of whether a human face is recognized in image data captured by the perception subsystem in a direction from which audio data has been captured by the audio subsystem,

a second indication of whether a gaze is detected in image data captured by the perception subsystem in a direction from which audio data has been captured by the audio subsystem, or

a third indication of whether mouth movement is detected in image data captured by the perception subsystem in a direction from which audio data has been captured by the audio subsystem.

11. The robot of claim 1 , wherein the one or more sensor subsystems comprise a touch detection subsystem and wherein the attention signals comprise a touch signal that indicates whether the robot is being touched by a user.

12. The robot of claim 1 , wherein the one or more sensor subsystems comprise a touch detection subsystem and a movement detection subsystem, and wherein the attention signals comprise a signal that indicates whether the robot has been picked up based on a first indication that the robot is being touched by a user according to the touch detection subsystem and a second indication that the robot has experienced movement according to the movement detection subsystem.

13. The robot of claim 1 , wherein the operations comprise:

maintaining a multidimensional emotion state of the robot, wherein the emotion state comprises respective values for each of a plurality of aspects; and

providing the multidimensional emotion state as input to the attention level classifier to generate the current attention level for the robot.

14. The robot of claim 1 , wherein the operations comprise:

maintaining a collection behavior update rules, wherein each behavior update rule is a mapping between a behavior, a behavior result, and an update to a particular aspect of the emotion state;

receiving a behavior result that represents an outcome of executing the determined behavior;

determining that a particular behavior update rule in the collection of behavior update rules includes the determined behavior and the behavior result; and

in response, updating the emotion state according to an update defined in the particular behavior update rule.

15. The robot of claim 1 , wherein the operations comprising:

maintaining a collection of attention level update rules, wherein each attention level update rule is a mapping between a first attention level, a second attention level, and an update to a particular aspect of the emotion state;

receiving a second attention level generated by the attention level classifier;

determining that a particular attention level update rule in the collection of attention level update rules includes the first attention level and the second attention level; and

in response, updating the emotion state according to an update defined in the particular attention level update rule.

16. An apparatus comprising:

one or more sensor subsystems that are configured to generate sensor inputs comprising respective representations of an environment;

one or more processors; and

one or more storage devices storing instructions that are operable, when executed by the one or more processors, to cause the apparatus to perform operations comprising:

computing a plurality of attention signals from the sensor inputs, wherein each attention signal corresponds to one factor of multiple factors for classifying an attention level using an attention level classifier according to the sensor inputs,

providing the plurality of attention signals as input to the attention level classifier to generate an attention level

determining that a user is paying attention to the apparatus based on the generated attention level;

in response, determining a behavior to execute based on the current attention level, wherein a behavior comprises one or more coordinated actions to be performed by the apparatus, and

executing the determined behavior including performing the one or more coordinated actions.

17. The apparatus of claim 16 , wherein executing the determined behavior comprises uploading a stream of received audio for natural language processing without having received a wake word.

18. The apparatus of claim 16 , wherein the attention level represents a predicted likelihood that a user is paying attention to the apparatus based on the sensor inputs.

19. The apparatus of claim 16 , wherein the one or more sensor subsystems comprise a perception subsystem configured to sense light from an environment of the robot, and

wherein the attention signals comprise:

a first indication of whether a human face is recognized in image data captured by the perception subsystem,

a second indication of whether a gaze is detected for at least a threshold amount of time in image data captured by the perception subsystem, or

a third indication of whether mouth movement is detected in image data captured by the perception subsystem.

20. A system comprising:

one or more sensor subsystems that are configured to generate sensor inputs comprising respective representations of an environment;

a robot comprising:

a body and one or more physically moveable components,

one or more processors, and

one or more storage devices storing instructions that are operable, when executed by the one or more processors, to cause the robot to perform operations comprising:

computing a plurality of attention signals from the sensor inputs, wherein each attention signal corresponds to one factor of multiple factors for classifying an attention level using an attention level classifier of the robot according to the sensor inputs,

providing the plurality of attention signals as input to the attention level classifier to generate an attention level

determining that a user is paying attention to the robot based on the generated attention level;

in response, determining a behavior to execute based on the current attention level, wherein a behavior comprises one or more coordinated actions to be performed by the robot, and

executing the determined behavior including performing the one or more coordinated actions.

21. The system of claim 20 , wherein the one or more sensor subsystems comprise a network-enabled user device configured to generate one or more of the sensor inputs.