Types∞

Overview∞

All types returned by agents are subclasses of Sample, ensuring a uniform interface and behavior across different agent outputs. Here are the main types associated with agents:

Sense∞

Sense objects are inputs to SensoryAgents. They provide raw data that the agents process into more structured forms.

Subclasses:

  • Image: Represents visual data.

  • Depth: Captures information about distances within a scene.

  • Audio: Contains sound data from the environment.

SensorReading∞

SensorReading objects are outputs from SensoryAgents, representing processed sensory data.

Subclasses:

  • 2DObjectPoses: Positions of objects in a 2D plane.

  • 2DBoundingBoxes: Bounding boxes around objects in a 2D plane.

  • 3DObjectPoses: Positions of objects in a 3D space.

  • 3DBoundingBoxes: Bounding boxes around objects in a 3D space.

  • Embedding: A numerical representation of data, often used in machine learning.

LanguageAction∞

A LanguageAction contains a list of instructions that can be executed. These instructions can be plain text commands or executable code, facilitating various language-based interactions.

Motion∞

Motion objects define movement parameters, including bounds (limits of movement), motion type (e.g., linear, rotational), and the reference frame (context or coordinate system for the motion).