C# ML代理不学习相对';简单';任务
我尝试创建一个简单的ML代理(ball)来学习如何向目标移动并与目标碰撞 不幸的是,代理似乎没有学习,只是在似乎是随机的位置上移动。5米步后,平均奖励保持在-1 对我做错了什么有什么建议吗 我的意见如下:C# ML代理不学习相对';简单';任务,c#,unity3d,ml-agent,C#,Unity3d,Ml Agent,我尝试创建一个简单的ML代理(ball)来学习如何向目标移动并与目标碰撞 不幸的是,代理似乎没有学习,只是在似乎是随机的位置上移动。5米步后,平均奖励保持在-1 对我做错了什么有什么建议吗 我的意见如下: /// <summary> /// Observations: /// 1: Distance to nearest target /// 3: Vector to nearest target /// 3: Target Position /// 3: Agent positi
/// <summary>
/// Observations:
/// 1: Distance to nearest target
/// 3: Vector to nearest target
/// 3: Target Position
/// 3: Agent position
/// 1: Agent Velocity X
/// 1: Agent Velocity Y
/// //12 observations in total
/// </summary>
/// <param name="sensor"></param>
public override void CollectObservations(VectorSensor sensor)
{
//If nearest Target is null, observe an empty array and return early
if (target == null)
{
sensor.AddObservation(new float[12]);
return;
}
float distanceToTarget = Vector3.Distance(target.transform.position, this.transform.position);
//Distance to nearest target (1 observervation)
sensor.AddObservation(distanceToTarget);
//Vector to nearest target (3 observations)
Vector3 toTarget = target.transform.position - this.transform.position;
sensor.AddObservation(toTarget.normalized);
//Target position
sensor.AddObservation(target.transform.localPosition);
//Current Position
sensor.AddObservation(this.transform.localPosition);
//Agent Velocities
sensor.AddObservation(rigidbody.velocity.x);
sensor.AddObservation(rigidbody.velocity.y);
}
奖励(全部在代理脚本上):
private void Update()
{
//如果特工从屏幕上掉下来,给予负面奖励,结束一集
if(此.transform.position.y<0)
{
增加奖励(-1.0f);
EndEpisode();
}
如果(目标!=null)
{
DrawLine(this.transform.position、target.transform.position、Color.green);
}
}
专用void OnCollisionEnter(碰撞)
{
//如果代理与目标发生冲突,则提供奖励
if(collidedObj.gameObject.CompareTag(“目标”))
{
增加奖励(1.0f);
摧毁(目标);
EndEpisode();
}
}
已接收公共覆盖无效OnAction(浮点[]矢量操作)
{
如果(!目标)
{
//放置并分配目标
envController.PlaceTarget();
target=envController.ProvideTarget();
}
矢量3控制信号=矢量3.0;
controlSignal.x=矢量作用[0];
controlSignal.z=矢量作用[1];
刚体.附加力(控制信号*移动速度,力模式.速度变化);
//在每一步都给予小小的负面奖励,以鼓励行动
如果(this.MaxStep>0)添加奖励(-1f/this.MaxStep);
}
您认为您的环境有多艰难?如果很少达到目标,代理将无法学习。在这种情况下,当代理朝着正确的方向行动时,您需要添加一些内在的奖励。这样,即使奖励很少,代理也可以学习
从您设计奖励的方式来看,奖励黑客也可能存在问题。如果代理无法找到目标以获得更大的奖励,最有效的方法是尽快从平台上摔下来,以免在每个时间段都受到小的惩罚。你在哪里奖励好的和坏的行为?现在添加到帖子中。谢谢,这很有意义。我在随机地点生成目标和代理,现在我意识到也许我需要通过课程学习来增加任务的难度。现在就开始尝试并找到一个关于如何做到这一点的教程@我很高兴,希望能帮上忙!如果您考虑所回答的问题,请接受左侧绿色复选标记的答案。
behaviors:
PlayerAgent:
trainer_type: ppo
hyperparameters:
batch_size: 512 #128
buffer_size: 2048
learning_rate: 3.0e-4
beta: 5.0e-4
epsilon: 0.2 #0.2
lambd: 0.99
num_epoch: 3 #3
learning_rate_schedule: linear
network_settings:
normalize: false
hidden_units: 32 #256
num_layers: 2
vis_encode_type: simple
reward_signals:
extrinsic:
gamma: 0.99
strength: 1.0
curiosity:
strength: 0.02
gamma: 0.99
encoding_size: 64
learning_rate: 3.0e-4
#keep_checkpoints: 5
#checkpoint_interval: 500000
max_steps: 5000000
time_horizon: 64
summary_freq: 10000
threaded: true
framework: tensorflow
private void Update()
{
//If Agent falls off the screen, give negative reward an end episode
if (this.transform.position.y < 0)
{
AddReward(-1.0f);
EndEpisode();
}
if(target != null)
{
Debug.DrawLine(this.transform.position, target.transform.position, Color.green);
}
}
private void OnCollisionEnter(Collision collidedObj)
{
//If agent collides with goal, provide reward
if (collidedObj.gameObject.CompareTag("Goal"))
{
AddReward(1.0f);
Destroy(target);
EndEpisode();
}
}
public override void OnActionReceived(float[] vectorAction)
{
if (!target)
{
//Place and assign the target
envController.PlaceTarget();
target = envController.ProvideTarget();
}
Vector3 controlSignal = Vector3.zero;
controlSignal.x = vectorAction[0];
controlSignal.z = vectorAction[1];
rigidbody.AddForce(controlSignal * moveSpeed, ForceMode.VelocityChange);
// Apply a tiny negative reward every step to encourage action
if (this.MaxStep > 0) AddReward(-1f / this.MaxStep);
}