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
 
 
 
 
 

43 行
1.7 KiB

using System.Collections.Generic;
using System;
namespace MLAgents
{
/// <summary>
/// Side channels provide an alternative mechanism of sending/receiving data from Unity
/// to Python that is outside of the traditional machine learning loop. ML-Agents provides
/// some specific implementations of side channels, but users can create their own.
/// </summary>
public abstract class SideChannel
{
// The list of messages (byte arrays) that need to be sent to Python via the communicator.
// Should only ever be read and cleared by a ICommunicator object.
internal List<byte[]> MessageQueue = new List<byte[]>();
/// <summary>
/// An int identifier for the SideChannel. Ensures that there is only ever one side channel
/// of each type. Ensure the Unity side channels will be linked to their Python equivalent.
/// </summary>
/// <returns> The integer identifier of the SideChannel.</returns>
public Guid ChannelId{
get;
protected set;
}
/// <summary>
/// Is called by the communicator every time a message is received from Python by the SideChannel.
/// Can be called multiple times per simulation step if multiple messages were sent.
/// </summary>
/// <param name="data"> the payload of the message.</param>
public abstract void OnMessageReceived(byte[] data);
/// <summary>
/// Queues a message to be sent to Python during the next simulation step.
/// </summary>
/// <param name="data"> The byte array of data to be sent to Python.</param>
protected void QueueMessageToSend(byte[] data)
{
MessageQueue.Add(data);
}
}
}