Unity 机器学习代理工具包 (ML-Agents) 是一个开源项目,它使游戏和模拟能够作为训练智能代理的环境。
您最多选择25个主题 主题必须以中文或者字母或数字开头,可以包含连字符 (-),并且长度不得超过35个字符
 
 
 
 
 

948 行
39 KiB

using System;
using System.Collections.Generic;
using UnityEngine;
using UnityEngine.Assertions;
using Unity.MLAgents.Sensors;
using UnityEngine.Profiling;
namespace Unity.MLAgents.Extensions.Sensors
{
/// <summary>
/// Grid-based sensor.
/// </summary>
[AddComponentMenu("ML Agents/Grid Sensor", (int)MenuGroup.Sensors)]
public class GridSensor : SensorComponent, ISensor, IBuiltInSensor
{
/// <summary>
/// Name of this grid sensor.
/// </summary>
public string Name;
//
// Main Parameters
//
/// <summary>
/// The width of each grid cell.
/// </summary>
[Header("Grid Sensor Settings")]
[Tooltip("The width of each grid cell")]
[Range(0.05f, 1000f)]
public float CellScaleX = 1f;
/// <summary>
/// The depth of each grid cell.
/// </summary>
[Tooltip("The depth of each grid cell")]
[Range(0.05f, 1000f)]
public float CellScaleZ = 1f;
/// <summary>
/// The width of the grid .
/// </summary>
[Tooltip("The width of the grid")]
[Range(2, 2000)]
public int GridNumSideX = 16;
/// <summary>
/// The depth of the grid .
/// </summary>
[Tooltip("The depth of the grid")]
[Range(2, 2000)]
public int GridNumSideZ = 16;
/// <summary>
/// The height of each grid cell. Changes how much of the vertical axis is observed by a cell.
/// </summary>
[Tooltip("The height of each grid cell. Changes how much of the vertical axis is observed by a cell")]
[Range(0.01f, 1000f)]
public float CellScaleY = 0.01f;
/// <summary>
/// Rotate the grid based on the direction the agent is facing.
/// </summary>
[Tooltip("Rotate the grid based on the direction the agent is facing")]
public bool RotateToAgent;
/// <summary>
/// Array holding the depth of each channel.
/// </summary>
[Tooltip("Array holding the depth of each channel")]
public int[] ChannelDepth;
/// <summary>
/// List of tags that are detected.
/// </summary>
[Tooltip("List of tags that are detected")]
public string[] DetectableObjects;
/// <summary>
/// The layer mask.
/// </summary>
[Tooltip("The layer mask")]
public LayerMask ObserveMask;
/// <summary>
/// Enum describing what kind of depth type the data should be organized as
/// </summary>
public enum GridDepthType { Channel, ChannelHot };
/// <summary>
/// The data layout that the grid should output.
/// </summary>
[Tooltip("The data layout that the grid should output")]
public GridDepthType gridDepthType = GridDepthType.Channel;
/// <summary>
/// The reference of the root of the agent. This is used to disambiguate objects with the same tag as the agent. Defaults to current GameObject.
/// </summary>
[Tooltip("The reference of the root of the agent. This is used to disambiguate objects with the same tag as the agent. Defaults to current GameObject")]
public GameObject rootReference;
[Header("Collider Buffer Properties")]
[Tooltip("The absolute max size of the Collider buffer used in the non-allocating Physics calls. In other words" +
" the Collider buffer will never grow beyond this number even if there are more Colliders in the Grid Cell.")]
public int MaxColliderBufferSize = 500;
[Tooltip(
"The Estimated Max Number of Colliders to expect per cell. This number is used to " +
"pre-allocate an array of Colliders in order to take advantage of the OverlapBoxNonAlloc " +
"Physics API. If the number of colliders found is >= InitialColliderBufferSize the array " +
"will be resized to double its current size. The hard coded absolute size is 500.")]
public int InitialColliderBufferSize = 4;
Collider[] m_ColliderBuffer;
float[] m_ChannelBuffer;
//
// Hidden Parameters
//
/// <summary>
/// The total number of observations per cell of the grid. Its equivalent to the "channel" on the outgoing tensor.
/// </summary>
[HideInInspector]
public int ObservationPerCell;
/// <summary>
/// The total number of observations that this GridSensor provides. It is the length of m_PerceptionBuffer.
/// </summary>
[HideInInspector]
public int NumberOfObservations;
/// <summary>
/// The offsets used to specify where within a cell's allotted data, certain observations will be inserted.
/// </summary>
[HideInInspector]
public int[] ChannelOffsets;
/// <summary>
/// The main storage of perceptual information.
/// </summary>
protected float[] m_PerceptionBuffer;
/// <summary>
/// The default value of the perceptionBuffer when using the ChannelHot DepthType. Used to reset the array/
/// </summary>
protected float[] m_ChannelHotDefaultPerceptionBuffer;
/// <summary>
/// Array of Colors needed in order to load the values of the perception buffer to a texture.
/// </summary>
protected Color[] m_PerceptionColors;
/// <summary>
/// Texture where the colors are written to so that they can be compressed in PNG format.
/// </summary>
protected Texture2D m_perceptionTexture2D;
//
// Utility Constants Calculated on Init
//
/// <summary>
/// Number of PNG formated images that are sent to python during training.
/// </summary>
private int NumImages;
/// <summary>
/// Number of relevant channels on the last image that is sent/
/// </summary>
private int NumChannelsOnLastImage;
/// <summary>
/// Radius of grid, used for normalizing the distance.
/// </summary>
protected float InverseSphereRadius;
/// <summary>
/// Total Number of cells (width*height)
/// </summary>
private int NumCells;
/// <summary>
/// Difference between GridNumSideZ and gridNumSideX.
/// </summary>
protected int DiffNumSideZX = 0;
/// <summary>
/// Offset used for calculating CellToPoint
/// </summary>
protected float OffsetGridNumSide = 7.5f; // (gridNumSideZ - 1) / 2;
/// <summary>
/// Half of the grid in the X direction
/// </summary>
private float HalfOfGridX;
/// <summary>
/// Half of the grid in the z direction
/// </summary>
private float HalfOfGridZ;
/// <summary>
/// Used in the PointToCell method to scale the x value to land in the calculated cell.
/// </summary>
private float PointToCellScalingX;
/// <summary>
/// Used in the PointToCell method to scale the y value to land in the calculated cell.
/// </summary>
private float PointToCellScalingZ;
/// <summary>
/// Bool if initialized or not.
/// </summary>
protected bool Initialized = false;
/// <summary>
/// Cached ObservationSpec
/// </summary>
private ObservationSpec m_ObservationSpec;
//
// Debug Parameters
//
/// <summary>
/// Array of Colors used for the grid gizmos.
/// </summary>
[Header("Debug Options")]
[Tooltip("Array of Colors used for the grid gizmos")]
public Color[] DebugColors;
/// <summary>
/// The height of the gizmos grid.
/// </summary>
[Tooltip("The height of the gizmos grid")]
public float GizmoYOffset = 0f;
/// <summary>
/// Whether to show gizmos or not.
/// </summary>
[Tooltip("Whether to show gizmos or not")]
public bool ShowGizmos = false;
public SensorCompressionType CompressionType = SensorCompressionType.PNG;
/// <summary>
/// Array of colors displaying the DebugColors for each cell in OnDrawGizmos. Only updated if ShowGizmos.
/// </summary>
protected Color[] CellActivity;
/// <summary>
/// Array of positions where each position is the center of a cell.
/// </summary>
private Vector3[] CellPoints;
/// <summary>
/// List representing the multiple compressed images of all of the grids
/// </summary>
private List<byte[]> compressedImgs;
/// <summary>
/// List representing the sizes of the multiple images so they can be properly reconstructed on the python side
/// </summary>
private List<byte[]> byteSizesBytesList;
private Color DebugDefaultColor = new Color(1f, 1f, 1f, 0.25f);
/// <inheritdoc/>
public override ISensor CreateSensor()
{
return this;
}
/// <summary>
/// Sets the parameters of the grid sensor
/// </summary>
/// <param name="detectableObjects">array of strings representing the tags to be detected by the sensor</param>
/// <param name="channelDepth">array of ints representing the depth of each channel</param>
/// <param name="gridDepthType">enum representing the GridDepthType of the sensor</param>
/// <param name="cellScaleX">float representing the X scaling of each cell</param>
/// <param name="cellScaleZ">float representing the Z scaling of each cell</param>
/// <param name="gridWidth">int representing the number of cells in the X direction. Width of the Grid</param>
/// <param name="gridHeight">int representing the number of cells in the Z direction. Height of the Grid</param>
/// <param name="observeMaskInt">int representing the layer mask to observe</param>
/// <param name="rotateToAgent">bool if true then the grid is rotated to the rotation of the transform the rootReference</param>
/// <param name="debugColors">array of colors corresponding the the tags in the detectableObjects array</param>
public virtual void SetParameters(string[] detectableObjects, int[] channelDepth, GridDepthType gridDepthType,
float cellScaleX, float cellScaleZ, int gridWidth, int gridHeight, int observeMaskInt, bool rotateToAgent, Color[] debugColors)
{
this.ObserveMask = observeMaskInt;
this.DetectableObjects = detectableObjects;
this.ChannelDepth = channelDepth;
this.gridDepthType = gridDepthType;
this.CellScaleX = cellScaleX;
this.CellScaleZ = cellScaleZ;
this.GridNumSideX = gridWidth;
this.GridNumSideZ = gridHeight;
this.RotateToAgent = rotateToAgent;
this.DiffNumSideZX = (GridNumSideZ - GridNumSideX);
this.OffsetGridNumSide = (GridNumSideZ - 1f) / 2f;
this.DebugColors = debugColors;
}
/// <summary>
/// Initializes the constant parameters used within the perceive method call
/// </summary>
public void InitGridParameters()
{
NumCells = GridNumSideX * GridNumSideZ;
float sphereRadiusX = (CellScaleX * GridNumSideX) / Mathf.Sqrt(2);
float sphereRadiusZ = (CellScaleZ * GridNumSideZ) / Mathf.Sqrt(2);
InverseSphereRadius = 1.0f / Mathf.Max(sphereRadiusX, sphereRadiusZ);
ChannelOffsets = new int[ChannelDepth.Length];
DiffNumSideZX = (GridNumSideZ - GridNumSideX);
OffsetGridNumSide = (GridNumSideZ - 1f) / 2f;
HalfOfGridX = CellScaleX * GridNumSideX / 2;
HalfOfGridZ = CellScaleZ * GridNumSideZ / 2;
PointToCellScalingX = GridNumSideX / (CellScaleX * GridNumSideX);
PointToCellScalingZ = GridNumSideZ / (CellScaleZ * GridNumSideZ);
}
/// <summary>
/// Initializes the constant parameters that are based on the Grid Depth Type
/// Sets the ObservationPerCell and the ChannelOffsets properties
/// </summary>
public virtual void InitDepthType()
{
switch (gridDepthType)
{
case GridDepthType.Channel:
ObservationPerCell = ChannelDepth.Length;
break;
case GridDepthType.ChannelHot:
ObservationPerCell = 0;
ChannelOffsets[ChannelOffsets.Length - 1] = 0;
for (int i = 1; i < ChannelDepth.Length; i++)
{
ChannelOffsets[i] = ChannelOffsets[i - 1] + ChannelDepth[i - 1];
}
for (int i = 0; i < ChannelDepth.Length; i++)
{
ObservationPerCell += ChannelDepth[i];
}
break;
}
// The maximum number of channels in the final output must be less than 255 * 3 because the "number of PNG images" to generate must fit in one byte
Assert.IsTrue(ObservationPerCell < (255 * 3), "The maximum number of channels per cell must be less than 255 * 3");
}
/// <summary>
/// Initializes the location of the CellPoints property
/// </summary>
private void InitCellPoints()
{
CellPoints = new Vector3[NumCells];
for (int i = 0; i < NumCells; i++)
{
CellPoints[i] = CellToPoint(i, false);
}
}
/// <summary>
/// Initializes the m_ChannelHotDefaultPerceptionBuffer with default data in the case that the grid depth type is ChannelHot
/// </summary>
public virtual void InitChannelHotDefaultPerceptionBuffer()
{
m_ChannelHotDefaultPerceptionBuffer = new float[ObservationPerCell];
for (int i = 0; i < ChannelDepth.Length; i++)
{
if (ChannelDepth[i] > 1)
{
m_ChannelHotDefaultPerceptionBuffer[ChannelOffsets[i]] = 1;
}
}
}
/// <summary>
/// Initializes the m_PerceptionBuffer as the main data storage property
/// Calculates the NumImages and NumChannelsOnLastImage that are used for serializing m_PerceptionBuffer
/// </summary>
public void InitPerceptionBuffer()
{
if (Application.isPlaying)
Initialized = true;
NumberOfObservations = ObservationPerCell * NumCells;
m_PerceptionBuffer = new float[NumberOfObservations];
if (gridDepthType == GridDepthType.ChannelHot)
{
InitChannelHotDefaultPerceptionBuffer();
}
m_PerceptionColors = new Color[NumCells];
NumImages = ObservationPerCell / 3;
NumChannelsOnLastImage = ObservationPerCell % 3;
if (NumChannelsOnLastImage == 0)
NumChannelsOnLastImage = 3;
else
NumImages += 1;
CellActivity = new Color[NumCells];
}
/// <summary>
/// Calls the initialization methods. Creates the data storing properties used to send the data
/// Establishes
/// </summary>
public virtual void Start()
{
InitGridParameters();
InitDepthType();
InitCellPoints();
InitPerceptionBuffer();
m_ColliderBuffer = new Collider[Math.Min(MaxColliderBufferSize, InitialColliderBufferSize)];
// Default root reference to current game object
if (rootReference == null)
rootReference = gameObject;
m_ObservationSpec = ObservationSpec.Visual(GridNumSideX, GridNumSideZ, ObservationPerCell);
compressedImgs = new List<byte[]>();
byteSizesBytesList = new List<byte[]>();
m_perceptionTexture2D = new Texture2D(GridNumSideX, GridNumSideZ, TextureFormat.RGB24, false);
}
/// <inheritdoc cref="ISensor.Reset"/>
void ISensor.Reset() { }
/// <summary>
/// Clears the perception buffer before loading in new data. If the gridDepthType is ChannelHot, then it initializes the
/// Reset() also reinits the cell activity array (for debug)
/// </summary>
public void ClearPerceptionBuffer()
{
if (m_PerceptionBuffer != null)
{
if (gridDepthType == GridDepthType.ChannelHot)
{
// Copy the default value to the array
for (int i = 0; i < NumCells; i++)
{
Array.Copy(m_ChannelHotDefaultPerceptionBuffer, 0, m_PerceptionBuffer, i * ObservationPerCell, ObservationPerCell);
}
}
else
{
Array.Clear(m_PerceptionBuffer, 0, m_PerceptionBuffer.Length);
}
}
else
{
m_PerceptionBuffer = new float[NumberOfObservations];
m_ColliderBuffer = new Collider[Math.Min(MaxColliderBufferSize, InitialColliderBufferSize)];
}
if (ShowGizmos)
{
// Ensure to init arrays if not yet assigned (for editor)
if (CellActivity == null)
CellActivity = new Color[NumCells];
// Assign the default color to the cell activities
for (int i = 0; i < NumCells; i++)
{
CellActivity[i] = DebugDefaultColor;
}
}
}
/// <inheritdoc/>
public string GetName()
{
return Name;
}
/// <inheritdoc/>
public virtual SensorCompressionType GetCompressionType()
{
return CompressionType;
}
/// <inheritdoc/>
public BuiltInSensorType GetBuiltInSensorType()
{
return BuiltInSensorType.GridSensor;
}
/// <summary>
/// GetCompressedObservation - Calls Perceive then puts the data stored on the perception buffer
/// onto the m_perceptionTexture2D to be converted to a byte array and returned
/// </summary>
/// <returns>byte[] containing the compressed observation of the grid observation</returns>
public byte[] GetCompressedObservation()
{
using (TimerStack.Instance.Scoped("GridSensor.GetCompressedObservation"))
{
Perceive(); // Fill the perception buffer with observed data
var allBytes = new List<byte>();
for (int i = 0; i < NumImages - 1; i++)
{
ChannelsToTexture(3 * i, 3);
allBytes.AddRange(m_perceptionTexture2D.EncodeToPNG());
}
ChannelsToTexture(3 * (NumImages - 1), NumChannelsOnLastImage);
allBytes.AddRange(m_perceptionTexture2D.EncodeToPNG());
return allBytes.ToArray();
}
}
/// <summary>
/// ChannelsToTexture - Takes the channel index and the numChannelsToAdd.
/// For each cell and for each channel to add, sets it to a value of the color specified for that cell.
/// All colors are then set to the perceptionTexture via SetPixels.
/// m_perceptionTexture2D can then be read as an image as it now contains all of the information that was
/// stored in the channels
/// </summary>
/// <param name="channelIndex"></param>
/// <param name="numChannelsToAdd"></param>
protected void ChannelsToTexture(int channelIndex, int numChannelsToAdd)
{
for (int i = 0; i < NumCells; i++)
{
for (int j = 0; j < numChannelsToAdd; j++)
{
m_PerceptionColors[i][j] = m_PerceptionBuffer[i * ObservationPerCell + channelIndex + j];
}
}
m_perceptionTexture2D.SetPixels(m_PerceptionColors);
}
/// <summary>
/// Perceive - Clears the buffers, calls overlap box on the actual cell (the actual perception part)
/// for all found colliders, LoadObjectData is called
/// at the end, Perceive returns the float array of the perceptions
/// </summary>
/// <returns>A float[] containing all of the information collected from the gridsensor</returns>
public float[] Perceive()
{
if (m_ColliderBuffer == null)
{
return Array.Empty<float>();
}
ClearPerceptionBuffer();
using (TimerStack.Instance.Scoped("GridSensor.Perceive"))
{
var halfCellScale = new Vector3(CellScaleX / 2f, CellScaleY, CellScaleZ / 2f);
for (var cellIndex = 0; cellIndex < NumCells; cellIndex++)
{
int numFound;
Vector3 cellCenter;
if (RotateToAgent)
{
Transform transform1;
cellCenter = (transform1 = transform).TransformPoint(CellPoints[cellIndex]);
numFound = BufferResizingOverlapBoxNonAlloc(cellCenter, halfCellScale, transform1.rotation);
}
else
{
cellCenter = transform.position + CellPoints[cellIndex];
numFound = BufferResizingOverlapBoxNonAlloc(cellCenter, halfCellScale, Quaternion.identity);
}
if (numFound > 0)
{
ParseColliders(m_ColliderBuffer, numFound, cellIndex, cellCenter);
}
}
}
return m_PerceptionBuffer;
}
/// <summary>
/// This method attempts to perform the Physics.OverlapBoxNonAlloc and will double the size of the Collider buffer
/// if the number of Colliders in the buffer after the call is equal to the length of the buffer.
/// </summary>
/// <param name="cellCenter"></param>
/// <param name="halfCellScale"></param>
/// <param name="rotation"></param>
/// <returns></returns>
int BufferResizingOverlapBoxNonAlloc(Vector3 cellCenter, Vector3 halfCellScale, Quaternion rotation)
{
int numFound;
// Since we can only get a fixed number of results, requery
// until we're sure we can hold them all (or until we hit the max size).
while (true)
{
numFound = Physics.OverlapBoxNonAlloc(cellCenter, halfCellScale, m_ColliderBuffer, rotation, ObserveMask);
if (numFound == m_ColliderBuffer.Length && m_ColliderBuffer.Length < MaxColliderBufferSize)
{
m_ColliderBuffer = new Collider[Math.Min(MaxColliderBufferSize, m_ColliderBuffer.Length * 2)];
InitialColliderBufferSize = m_ColliderBuffer.Length;
}
else
{
break;
}
}
return numFound;
}
/// <summary>
/// Parses the array of colliders found within a cell. Finds the closest gameobject to the agent root reference within the cell
/// </summary>
/// <param name="foundColliders">Array of the colliders found within the cell</param>
/// <param name="numFound">Number of colliders found.</param>
/// <param name="cellIndex">The index of the cell</param>
/// <param name="cellCenter">The center position of the cell</param>
protected virtual void ParseColliders(Collider[] foundColliders, int numFound, int cellIndex, Vector3 cellCenter)
{
Profiler.BeginSample("GridSensor.ParseColliders");
GameObject closestColliderGo = null;
var minDistanceSquared = float.MaxValue;
for (var i = 0; i < numFound; i++)
{
var currentColliderGo = foundColliders[i].gameObject;
// Continue if the current collider go is the root reference
if (ReferenceEquals(currentColliderGo, rootReference))
continue;
var closestColliderPoint = foundColliders[i].ClosestPointOnBounds(cellCenter);
var currentDistanceSquared = (closestColliderPoint - rootReference.transform.position).sqrMagnitude;
// Checks if our colliders contain a detectable object
var index = -1;
for (var ii = 0; ii < DetectableObjects.Length; ii++)
{
if (currentColliderGo.CompareTag(DetectableObjects[ii]))
{
index = ii;
break;
}
}
if (index > -1 && currentDistanceSquared < minDistanceSquared)
{
minDistanceSquared = currentDistanceSquared;
closestColliderGo = currentColliderGo;
}
}
if (!ReferenceEquals(closestColliderGo, null))
LoadObjectData(closestColliderGo, cellIndex, (float)Math.Sqrt(minDistanceSquared) * InverseSphereRadius);
Profiler.EndSample();
}
/// <summary>
/// GetObjectData - returns an array of values that represent the game object
/// This is one of the few methods that one may need to override to get their required functionality
/// For instance, if one wants specific information about the current gameobject, they can use this method
/// to extract it and then return it in an array format.
/// </summary>
/// <returns>
/// A float[] containing the data that holds the representative information of the passed in gameObject
/// </returns>
/// <param name="currentColliderGo">The game object that was found colliding with a certain cell</param>
/// <param name="typeIndex">The index of the type (tag) of the gameObject.
/// (e.g., if this GameObject had the 3rd tag out of 4, type_index would be 2.0f)</param>
/// <param name="normalizedDistance">A float between 0 and 1 describing the ratio of
/// the distance currentColliderGo is compared to the edge of the gridsensor</param>
/// <example>
/// Here is an example of extenind GetObjectData to include information about a potential Rigidbody:
/// <code>
/// protected override float[] GetObjectData(GameObject currentColliderGo,
/// float type_index, float normalized_distance)
/// {
/// float[] channelValues = new float[ChannelDepth.Length]; // ChannelDepth.Length = 4 in this example
/// channelValues[0] = type_index;
/// Rigidbody goRb = currentColliderGo.GetComponent&lt;Rigidbody&gt;();
/// if (goRb != null)
/// {
/// channelValues[1] = goRb.velocity.x;
/// channelValues[2] = goRb.velocity.y;
/// channelValues[3] = goRb.velocity.z;
/// }
/// return channelValues;
/// }
/// </code>
/// </example>
protected virtual float[] GetObjectData(GameObject currentColliderGo, float typeIndex, float normalizedDistance)
{
if (m_ChannelBuffer == null)
{
m_ChannelBuffer = new float[ChannelDepth.Length];
}
Array.Clear(m_ChannelBuffer, 0, m_ChannelBuffer.Length);
m_ChannelBuffer[0] = typeIndex;
return m_ChannelBuffer;
}
/// <summary>
/// Runs basic validation assertions to check that the values can be normalized
/// </summary>
/// <param name="channelValues">The values to be validated</param>
/// <param name="currentColliderGo">The gameobject used for better error messages</param>
protected virtual void ValidateValues(float[] channelValues, GameObject currentColliderGo)
{
for (int j = 0; j < channelValues.Length; j++)
{
if (channelValues[j] < 0)
throw new UnityAgentsException("Expected ChannelValue[" + j + "] for " + currentColliderGo.name + " to be non-negative, was " + channelValues[j]);
if (channelValues[j] > ChannelDepth[j])
throw new UnityAgentsException("Expected ChannelValue[" + j + "] for " + currentColliderGo.name + " to be less than ChannelDepth[" + j + "] (" + ChannelDepth[j] + "), was " + channelValues[j]);
}
}
/// <summary>
/// LoadObjectData - If the GameObject matches a tag, GetObjectData is called to extract the data from the GameObject
/// then the data is transformed based on the GridDepthType of the gridsensor.
/// Further documetation on the GridDepthType can be found below
/// </summary>
/// <param name="currentColliderGo">The game object that was found colliding with a certain cell</param>
/// <param name="cellIndex">The index of the current cell</param>
/// <param name="normalizedDistance">A float between 0 and 1 describing the ratio of
/// the distance currentColliderGo is compared to the edge of the gridsensor</param>
protected virtual void LoadObjectData(GameObject currentColliderGo, int cellIndex, float normalizedDistance)
{
Profiler.BeginSample("GridSensor.LoadObjectData");
var channelHotVals = new ArraySegment<float>(m_PerceptionBuffer, cellIndex * ObservationPerCell, ObservationPerCell);
for (var i = 0; i < DetectableObjects.Length; i++)
{
for (var ii = 0; ii < channelHotVals.Count; ii++)
{
m_PerceptionBuffer[channelHotVals.Offset + ii] = 0f;
}
if (!ReferenceEquals(currentColliderGo, null) && currentColliderGo.CompareTag(DetectableObjects[i]))
{
// TODO: Create the array already then set the values using "out" in GetObjectData
// Using i+1 as the type index as "0" represents "empty"
float[] channelValues = GetObjectData(currentColliderGo, (float)i + 1, normalizedDistance);
ValidateValues(channelValues, currentColliderGo);
if (ShowGizmos)
{
Color debugRayColor = Color.white;
if (DebugColors.Length > 0)
{
debugRayColor = DebugColors[i];
}
CellActivity[cellIndex] = new Color(debugRayColor.r, debugRayColor.g, debugRayColor.b, .5f);
}
switch (gridDepthType)
{
case GridDepthType.Channel:
{
// The observations are "channel based" so each grid is WxHxC where C is the number of channels
// This typically means that each channel value is normalized between 0 and 1
// If channelDepth is 1, the value is assumed normalized, else the value is normalized by the channelDepth
// The channels are then stored consecutively in PerceptionBuffer.
// NOTE: This is the only grid type that uses floating point values
// For example, if a cell contains the 3rd type of 5 possible on the 2nd team of 3 possible teams:
// channelValues = {2, 1}
// ObservationPerCell = channelValues.Length
// channelValues = {2f/5f, 1f/3f} = {.4, .33..}
// Array.Copy(channelValues, 0, PerceptionBuffer, cell_id*ObservationPerCell, ObservationPerCell);
for (int j = 0; j < channelValues.Length; j++)
{
channelValues[j] /= ChannelDepth[j];
}
Array.Copy(channelValues, 0, m_PerceptionBuffer, cellIndex * ObservationPerCell, ObservationPerCell);
break;
}
case GridDepthType.ChannelHot:
{
// The observations are "channel hot" so each grid is WxHxD where D is the sum of all of the channel depths
// The opposite of the "channel based" case, the channel values are represented as one hot vector per channel and then concatenated together
// Thus channelDepth is assumed to be greater than 1.
// For example, if a cell contains the 3rd type of 5 possible on the 2nd team of 3 possible teams,
// channelValues = {2, 1}
// channelOffsets = {5, 3}
// ObservationPerCell = 5 + 3 = 8
// channelHotVals = {0, 0, 1, 0, 0, 0, 1, 0}
// Array.Copy(channelHotVals, 0, PerceptionBuffer, cell_id*ObservationPerCell, ObservationPerCell);
for (int j = 0; j < channelValues.Length; j++)
{
if (ChannelDepth[j] > 1)
{
m_PerceptionBuffer[channelHotVals.Offset + (int)channelValues[j] + ChannelOffsets[j]] = 1f;
}
else
{
m_PerceptionBuffer[channelHotVals.Offset + ChannelOffsets[j]] = channelValues[j];
}
}
break;
}
}
break;
}
}
Profiler.EndSample();
}
/// <summary>Converts the index of the cell to the 3D point (y is zero)</summary>
/// <returns>Vector3 of the position of the center of the cell</returns>
/// <param name="cell">The index of the cell</param>
/// <param name="shouldTransformPoint">Bool weather to transform the point to the current transform</param>
protected Vector3 CellToPoint(int cell, bool shouldTransformPoint = true)
{
float x = (cell % GridNumSideZ - OffsetGridNumSide) * CellScaleX;
float z = (cell / GridNumSideZ - OffsetGridNumSide) * CellScaleZ - DiffNumSideZX;
if (shouldTransformPoint)
return transform.TransformPoint(new Vector3(x, 0, z));
return new Vector3(x, 0, z);
}
/// <summary>Finds the cell in which the given global point falls</summary>
/// <returns>
/// The index of the cell in which the global point falls or -1 if the point does not fall into a cell
/// </returns>
/// <param name="globalPoint">The 3D point in global space</param>
public int PointToCell(Vector3 globalPoint)
{
Vector3 point = transform.InverseTransformPoint(globalPoint);
if (point.x < -HalfOfGridX || point.x > HalfOfGridX || point.z < -HalfOfGridZ || point.z > HalfOfGridZ)
return -1;
float x = point.x + HalfOfGridX;
float z = point.z + HalfOfGridZ;
int _x = (int)Mathf.Floor(x * PointToCellScalingX);
int _z = (int)Mathf.Floor(z * PointToCellScalingZ);
return GridNumSideX * _z + _x;
}
/// <summary>Copies the data from one cell to another</summary>
/// <param name="fromCellID">index of the cell to copy from</param>
/// <param name="toCellID">index of the cell to copy into</param>
protected void CopyCellData(int fromCellID, int toCellID)
{
Array.Copy(m_PerceptionBuffer,
fromCellID * ObservationPerCell,
m_PerceptionBuffer,
toCellID * ObservationPerCell,
ObservationPerCell);
if (ShowGizmos)
CellActivity[toCellID] = CellActivity[fromCellID];
}
void OnDrawGizmos()
{
if (ShowGizmos)
{
if (Application.isEditor && !Application.isPlaying)
Start();
Perceive();
var scale = new Vector3(CellScaleX, 1, CellScaleZ);
var offset = new Vector3(0, GizmoYOffset, 0);
var oldGizmoMatrix = Gizmos.matrix;
for (var i = 0; i < NumCells; i++)
{
Matrix4x4 cubeTransform;
if (RotateToAgent)
{
cubeTransform = Matrix4x4.TRS(CellToPoint(i) + offset, transform.rotation, scale);
}
else
{
cubeTransform = Matrix4x4.TRS(CellToPoint(i, false) + transform.position + offset, Quaternion.identity, scale);
}
Gizmos.matrix = oldGizmoMatrix * cubeTransform;
Gizmos.color = CellActivity[i];
Gizmos.DrawCube(Vector3.zero, Vector3.one);
}
Gizmos.matrix = oldGizmoMatrix;
if (Application.isEditor && !Application.isPlaying)
DestroyImmediate(m_perceptionTexture2D);
}
}
/// <inheritdoc/>
void ISensor.Update()
{
using (TimerStack.Instance.Scoped("GridSensor.Update"))
{
Perceive();
}
}
/// <summary>Gets the observation shape</summary>
/// <returns>int[] of the observation shape</returns>
public ObservationSpec GetObservationSpec()
{
// Lazy update
var shape = m_ObservationSpec.Shape;
if (shape[0] != GridNumSideX || shape[1] != GridNumSideZ || shape[2] != ObservationPerCell)
{
m_ObservationSpec = ObservationSpec.Visual(GridNumSideX, GridNumSideZ, ObservationPerCell);
}
return m_ObservationSpec;
}
/// <inheritdoc/>
public override int[] GetObservationShape()
{
var shape = m_ObservationSpec.Shape;
return new int[] { shape[0], shape[1], shape[2] };
}
/// <inheritdoc/>
public int Write(ObservationWriter writer)
{
using (TimerStack.Instance.Scoped("GridSensor.WriteToTensor"))
{
int index = 0;
for (var h = GridNumSideZ - 1; h >= 0; h--) // height
{
for (var w = 0; w < GridNumSideX; w++) // width
{
for (var d = 0; d < ObservationPerCell; d++) // depth
{
writer[h, w, d] = m_PerceptionBuffer[index];
index++;
}
}
}
return index;
}
}
}
}