Simulation

Command Line Arguments

The simulation can be configured by applying the following command line arguments. Unspecified arguments will be defaulted (default values in []).

The simulation supports the following arguments:

• --logLevel [0] : Logging level between 0 (minimum) and 5 (maximum - debug core)

• --logFile [opSimulation.log]* : Name of the log file

• --lib [modules] : Path of the libraries (relative or absolute)

• --configs [configs] : Path for writing outputs (relative or absolute)

• --results [results] : Path for writing outputs (relative or absolute)

Scheduler

The Scheduler handles the sequence of agent based and common tasks before, during and after simulation.

Executing phases

Each simulation run splits up into 3 phases: Bootstrap tasks, Stepping and Finalize tasks.

1. Initial task are grouped within the bootstrap tasks and are executed first and only once. This includes updating the Observation modules and triggering all PreRun Spawners.

2. After this Spawning, PreAgent, NonRecurring, Recurring and Syncronize Tasks are processed for each timestep. First all Runtime Spawners are triggered. Whenever a new agent is spawned, all its components are parsed to Trigger and Update tasks, grouped and sorted by execution time as non-recurring and recurring tasks. Next all PreAgents task are executed, which includes publishing the world state to the DataBuffer in triggering all EventDetectors and Manipulators Non-recurring tasks are executed just once and will be deleted afterwards (init flag in ComponentRepository). Right after execution of all agent-based tasks, the Syncronize phase synchronizes all changes to the world as world state for the next timestep.

3. The last phase is finalizing task, executed when an end condition is reached, such as end of simulation duration. Within this phase all EventDetectors and Manipulators are triggered and the Observation modules are updated one last time.

Task type description

The scheduler handles 8 different task types:

• Spawning - triggers agent spawning

• EventDetector - execute event detector

• Manipulator - execute manipulator

• Observation - update observation modules

• Trigger - execute trigger function of connected component

• Update - execute update output of connected component and resulting input

• SyncWorld - update world

The following table gives an overview to all phases.

Phase	Changeable	Task types	Notes
Bootstrap	no	Spawning (PreRun), Observation
Spawning	no	Spawning (Runtime), Observation
PreAgent	no	SyncWorld (publishing), EventDetector, Manipulator
Non recurring	yes	Trigger, Update	Only components defined as “Init”
Recurring	yes	Trigger, Update	Non Init components
Syncronize	no	SyncWorld	Update the state of the agents and the virtual world (e.g. due to agent movements).
Finalize	no	EventDetector, Manipulator, Observation

Task priorities

The order of the tasks within one phase depends on component priority and task type. All tasks for a given timestamp are ordered by the SchedulerTasks priority, and divided into the aformentioned phases.

The following table shows the priorities for the non-component task types:

Task type	Priority	Notes
Spawning	4	Highest priority, nothing can happen before any agent is instantiated.
EventDetector	3
Manipulator	2	The manipulator task uses the event detector tasks as input. Therefore it is called afterwards.
SyncGlobalData	1
Observation	0	Observation tasks have to be execute at the end.

The priority of trigger and update tasks depend on the priority within the SystemConfig. They are independent of non-component priorities. Since trigger tasks prepare output signals, output tasks are right called after the corresponding trigger tasks.