add team reward

Project/Assets/ML-Agents/Examples/PushBlock/Scripts/PushBlockTeamManager.cs

 using System.Collections.Generic;
 using Unity.MLAgents;
+using System.Linq;
-    List<Agent> m_AgentList = new List<Agent> { };
+    Dictionary<Agent, bool> m_AgentDone = new Dictionary<Agent, bool> { };
-        m_AgentList.Add(agent);
+        m_AgentDone[agent] = false;
+    }
+
+    public override void OnAgentDone(Agent agent, Agent.DoneReason doneReason, List<ISensor> sensors)
+    {
+        m_AgentDone[agent] = true;
+    }
+
+    public void OnTeamDone()
+    {
+        foreach (var agent in m_AgentDone.Keys.ToList())
+        {
+            if (m_AgentDone[agent])
+            {
+                agent.SendDoneToTrainer();
+                m_AgentDone[agent] = false;
+            }
+        }
+    }
+
+    public void AddTeamReward(float reward)
+    {
+        foreach (var agent in m_AgentDone.Keys)
+        {
+            if (m_AgentDone[agent])
+            {
+                agent.AddRewardAfterDeath(reward);
+            }
+            else
+            {
+                agent.AddReward(reward);
+            }
+        }
     }
 }

Project/Assets/ML-Agents/Examples/PushBlock/Scripts/ZombiePushBlockDeathEnvController.cs

     public bool UseRandomBlockRotation = true;
     public bool UseRandomBlockPosition = true;
     public bool UseTeamManager = true;
-    // public bool UseTeamReward = true;
+    public bool UseTeamReward = true;
     // public bool DiscountTeamReward = true;
     PushBlockSettings m_PushBlockSettings;
 
         // Initialize TeamManager
         if (UseTeamManager)
         {
-            m_TeamManager = new BaseTeamManager();
+            if (UseTeamReward)
+            {
+                m_TeamManager = new PushBlockTeamManager();
+            }
+            else
+            {
+                m_TeamManager = new BaseTeamManager();
+            }
         }
 
         foreach (var item in AgentsList)
     public void ScoredAGoal(Collider col, float score)
     {
         // Give Agent Rewards
-        foreach (var item in AgentsList)
+        if (UseTeamManager && UseTeamReward)
+        {
+            var pushManager = (PushBlockTeamManager)m_TeamManager;
+            pushManager.AddTeamReward(score);
+        }
+        else
-            if (item.Agent.gameObject.activeInHierarchy)
+            foreach (var item in AgentsList)
-                item.Agent.AddReward(score);
+                if (item.Agent.gameObject.activeInHierarchy)
+                {
+                    item.Agent.AddReward(score);
+                }
             }
         }
 
     public void ZombieTouchedBlock()
     {
         //Give Agent Rewards
-        foreach (var item in AgentsList)
+        if (UseTeamManager && UseTeamReward)
-            if (item.Agent.gameObject.activeInHierarchy)
+            var pushManager = (PushBlockTeamManager)m_TeamManager;
+            pushManager.AddTeamReward(-1);
+        }
+        else
+        {
+            foreach (var item in AgentsList)
-                item.Agent.AddReward(-1);
+                if (item.Agent.gameObject.activeInHierarchy)
+                {
+                    item.Agent.AddReward(-1);
+                }
             }
         }
 
                 return;
             }
             item.Agent.gameObject.SetActive(false);
+        }
+
+        if (UseTeamManager && UseTeamReward)
+        {
+            var pushManager = (PushBlockTeamManager)m_TeamManager;
+            pushManager.OnTeamDone();
         }
 
         // Reset Agents

com.unity.ml-agents/Runtime/Agent.cs

             m_CumulativeReward += increment;
         }
 
+        public void AddRewardAfterDeath(float increment)
+        {
+            m_Info.reward += increment;
+        }
+
         /// <summary>
         /// Retrieves the episode reward for the Agent.
         /// </summary>

config/ppo/PushBlockZombieTeamReward.yaml

+behaviors:
+  PushBlock:
+    trainer_type: ppo
+    hyperparameters:
+      batch_size: 128
+      buffer_size: 2048
+      learning_rate: 0.0003
+      beta: 0.01
+      epsilon: 0.2
+      lambd: 0.95
+      num_epoch: 3
+      learning_rate_schedule: linear
+    network_settings:
+      normalize: false
+      hidden_units: 256
+      num_layers: 2
+      vis_encode_type: simple
+    reward_signals:
+      extrinsic:
+        gamma: 0.99
+        strength: 1.0
+    keep_checkpoints: 5
+    max_steps: 15000000 #2000000
+    time_horizon: 64
+    summary_freq: 60000
+    threaded: true
+env_settings:
+    num_envs: 3