Class GridSensorBase

Grid-based sensor.

Inheritance
Object
GridSensorBase
OneHotGridSensor
Inherited Members
Namespace: Unity.MLAgents.Sensors
Syntax
public class GridSensorBase : ISensor, IBuiltInSensor, IDisposable

Constructors

GridSensorBase(String, Vector3, Vector3Int, String[], SensorCompressionType)

Create a GridSensorBase with the specified configuration.

Declaration
public GridSensorBase(string name, Vector3 cellScale, Vector3Int gridSize, string[] detectableTags, SensorCompressionType compression)
Parameters
Type Name Description
String name

The sensor name
Vector3 cellScale

The scale of each cell in the grid
Vector3Int gridSize

Number of cells on each side of the grid
String[] detectableTags

Tags to be detected by the sensor
SensorCompressionType compression

Compression type

Properties

CompressionType

The compression type used by the sensor.

Declaration
public SensorCompressionType CompressionType { get; set; }
Property Value
Type Description
SensorCompressionType

DetectableTags

The tags which the sensor dectects.

Declaration
protected string[] DetectableTags { get; }
Property Value
Type Description
String[]

Methods

Dispose()

Clean up the internal objects.

Declaration
public void Dispose()
Implements

GetBuiltInSensorType()

Declaration
public BuiltInSensorType GetBuiltInSensorType()
Returns
Type Description
BuiltInSensorType

GetCellObservationSize()

Get the observation size for each cell. This will be the size of dataBuffer for GetObjectData(GameObject, Int32, Single[]). If overriding GetObjectData(GameObject, Int32, Single[]), override this method as well to the custom observation size.

Declaration
protected virtual int GetCellObservationSize()
Returns
Type Description
Int32

The observation size of each cell.

GetCompressedObservation()

Return a compressed representation of the observation. For small observations, this should generally not be implemented. However, compressing large observations (such as visual results) can significantly improve model training time.

Declaration
public byte[] GetCompressedObservation()
Returns
Type Description
Byte[]

Compressed observation.
Implements

GetCompressionSpec()

Return information on the compression type being used. If no compression is used, return Default().

Declaration
public CompressionSpec GetCompressionSpec()
Returns
Type Description
CompressionSpec

An object describing the compression used by the sensor.
Implements

GetName()

Get the name of the sensor. This is used to ensure deterministic sorting of the sensors on an Agent, so the naming must be consistent across all sensors and agents.

Declaration
public string GetName()
Returns
Type Description
String

The name of the sensor.
Implements

GetObjectData(GameObject, Int32, Single[])

Get the observation values of the detected game object. Default is to record the detected tag index.

This method can be overridden to encode the observation differently or get custom data from the object. When overriding this method, GetCellObservationSize() and IsDataNormalized() might also need to change accordingly.

Declaration
protected virtual void GetObjectData(GameObject detectedObject, int tagIndex, float[] dataBuffer)
Parameters
Type Name Description
GameObject detectedObject

The game object that was detected within a certain cell
Int32 tagIndex

The index of the detectedObject's tag in the DetectableObjects list
Single[] dataBuffer

The buffer to write the observation values. The buffer size is configured by GetCellObservationSize().
Examples

Here is an example of overriding GetObjectData to get the velocity of a potential Rigidbody:

  protected override void GetObjectData(GameObject detectedObject, int tagIndex, float[] dataBuffer)
  {
      if (tagIndex == Array.IndexOf(DetectableTags, "RigidBodyObject"))
      {
          Rigidbody rigidbody = detectedObject.GetComponent<Rigidbody>();
          dataBuffer[0] = rigidbody.velocity.x;
          dataBuffer[1] = rigidbody.velocity.y;
          dataBuffer[2] = rigidbody.velocity.z;
      }
  }

GetObservationSpec()

Returns a description of the observations that will be generated by the sensor. See ObservationSpec for more details, and helper methods to create one.

Declaration
public ObservationSpec GetObservationSpec()
Returns
Type Description
ObservationSpec

An object describing the observation.
Implements

GetProcessCollidersMethod()

Whether to process all detected colliders in a cell. Default to false and only use the one closest to the agent. If overriding GetObjectData(GameObject, Int32, Single[]), consider override this method when needed.

Declaration
protected virtual ProcessCollidersMethod GetProcessCollidersMethod()
Returns
Type Description
ProcessCollidersMethod

Bool value indicating whether to process all detected colliders in a cell.

IsDataNormalized()

Whether the data is normalized within [0, 1]. The sensor can only use PNG compression if the data is normailzed. If overriding GetObjectData(GameObject, Int32, Single[]), override this method as well according to the custom observation values.

Declaration
protected virtual bool IsDataNormalized()
Returns
Type Description
Boolean

Bool value indicating whether data is normalized.

Reset()

Resets the internal state of the sensor. This is called at the end of an Agent's episode. Most implementations can leave this empty.

Declaration
public void Reset()
Implements

ResetPerceptionBuffer()

Clears the perception buffer before loading in new data.

Declaration
public void ResetPerceptionBuffer()

Update()

Update any internal state of the sensor. This is called once per each agent step.

Declaration
public void Update()
Implements

Write(ObservationWriter)

Write the observation data directly to the ObservationWriter. Note that this (and GetCompressedObservation()) may be called multiple times per agent step, so should not mutate any internal state.

Declaration
public int Write(ObservationWriter writer)
Parameters
Type Name Description
ObservationWriter writer

Where the observations will be written to.
Returns
Type Description
Int32

The number of elements written.
Implements

Extension Methods