Struct ObservationSpec

A description of the observations that an ISensor produces. This includes the size of the observation, the properties of each dimension, and how the observation should be used for training.

Inherited Members

ValueType.Equals(object)

ValueType.GetHashCode()

ValueType.ToString()

object.Equals(object, object)

object.GetType()

object.ReferenceEquals(object, object)

Namespace: Unity.MLAgents.Sensors

Assembly: Unity.ML-Agents.dll

Syntax

public struct ObservationSpec

Constructors

ObservationSpec(InplaceArray<int>, InplaceArray<DimensionProperty>, ObservationType)

Create a general ObservationSpec from the shape, dimension properties, and observation type.

Declaration

public ObservationSpec(InplaceArray<int> shape, InplaceArray<DimensionProperty> dimensionProperties, ObservationType observationType = ObservationType.Default)

Parameters

Type	Name	Description
InplaceArray<int>	shape	Shape
InplaceArray<DimensionProperty>	dimensionProperties	Dimension properties
ObservationType	observationType	Observation type

Remarks

Note that not all combinations of DimensionProperty may be supported by the trainer. shape and dimensionProperties must have the same size.

Exceptions

Type	Condition
UnityAgentsException	Unity agents exception

Properties

DimensionProperties

The properties of each dimensions of the observation. The length of the array must be equal to the rank of the observation tensor.

Declaration

public InplaceArray<DimensionProperty> DimensionProperties { get; }

Property Value

Type	Description
InplaceArray<DimensionProperty>

Remarks

It is generally recommended to use default values provided by helper functions, as not all combinations of DimensionProperty may be supported by the trainer.

ObservationType

The type of the observation, e.g. whether they are generic or help determine the goal for the Agent.

Declaration

public ObservationType ObservationType { get; }

Property Value

Type	Description
ObservationType

Rank

The number of dimensions of the observation.

Declaration

public int Rank { get; }

Property Value

Type	Description
int

Shape

The size of the observations that will be generated. For example, a sensor that observes the velocity of a rigid body (in 3D) would use [3]. A sensor that returns an RGB image would use [Height, Width, 3].

Declaration

public InplaceArray<int> Shape { get; }

Property Value

Type	Description
InplaceArray<int>

Methods

VariableLength(int, int)

Construct an ObservationSpec for variable-length observations.

Declaration

public static ObservationSpec VariableLength(int obsSize, int maxNumObs)

Parameters

Type	Name	Description
int	obsSize	Observation size
int	maxNumObs	Max number of observations

Returns

Type	Description
ObservationSpec	ObservationSpec for variable-length observations.

Vector(int, ObservationType)

Construct an ObservationSpec for 1-D observations of the requested length.

Declaration

public static ObservationSpec Vector(int length, ObservationType obsType = ObservationType.Default)

Parameters

Type	Name	Description
int	length	Length
ObservationType	obsType	Observation type

Returns

Type	Description
ObservationSpec	ObservationSpec for 1-D observations of the requested length.

Visual(int, int, int, ObservationType)

Construct an ObservationSpec for visual-like observations, e.g. observations with a height, width, and possible multiple channels.

Declaration

public static ObservationSpec Visual(int channels, int height, int width, ObservationType obsType = ObservationType.Default)

Parameters

Type	Name	Description
int	channels	Channels
int	height	Height
int	width	Width
ObservationType	obsType	Observation type

Returns

Type	Description
ObservationSpec	ObservationSpec for visual-like observations