Spawning and Removing Agents and Entities

To create instances of agent types and registering them at the runtime system to process the Tick(), two concepts are available:

Creating and Registering single object using RegisterAgent

Individual instances can be instantiated directly using new constructor expression. The instance must be registered with the runtime system by calling the RegisterAgent handle given through the InitLayer method of the ILayer implementation.

public bool InitLayer(TInitData layerInitData, 
    RegisterAgent registerAgentHandle, UnregisterAgent unregisterAgentHandle) 
{
    var layer = this;
    var agent = new MyAgentType();

    registerAgent(layer, agent);
    return true;
}

Optionally the RegisterAgent handle allows to set and override the execution frequency (for more see here) to specify after how many ticks the agent should be called again. Otherwise the registry uses the settings from the simulation configuration config.json.

If an output has been specified using output, the object is managed accordingly in the output as well automatically.

Mass Generation using AgentManager

The AgentManager is a service component that can be used to initialize agent quantities from the input.

The AgentManager is a static class that can be referenced directly and offers two methods:

• CreateAgents<Type>: This method creates a set of agents of the specified type but does not register them, so that their Tick() methods are called and their versions are managed in the output.
• SpawnAgents<Type>: This method creates and registers the set of agents of the specified type.

Spawning of agents is usually done within the InitLayer method of an ILayer implementation.

public bool InitLayer(
    TInitData layerInitData,
    RegisterAgent registerAgentHandle, 
    UnregisterAgent unregisterAgentHandle) 
{
    var layer = this;

    AgentManager.SpawnAgents<MyOtherAgent>(
        layerInitData.AgentInitConfigs.First(config => config.Name == "MyOtherAgent"), 
        registerAgentHandle, unregisterAgentHandle, 
        new List<ILayer> {layer});

    return true;
}

Since the agent interacts with the system via its layer, it is necessary that the AgentManager is given its own layer instance as dependency.

In the agent this dependent layer can be referenced as a property which is resolved by the system via PropertyInjection, for example as follows:

public class MyAgentType : AbstractAgent 
{
    [PropertyDescription]
    public MyLayerType Layer { get; set; }

    public void override Tick() {
        // access your layer here (e.g., Layer.Environment.Explore(...))
    }
}

Alternatively, dependent layer can be defined directly as constructor parameter as showed below:

danger

Because of the testing of your own model, it is recommended to keep the required constructor parameters as low as possible.

public class MyAgentType : AbstractAgent 
{
    public MyLayerType Layer { get; set; }

    public MyAgentType(MyLayerType layer) 
    {
        Layer = layer;
    }

    public void override Tick() {
        // access your layer here (e.g., Layer.Environment.Explore(...))
    }
}

Removing and unregister an object from simulation