Using Multiple Levels
One of the prominent features that we provide with the SWARM Platform is the ability to have multiple levels of an environment, allowing you to have multiple test cases to easily run in an automated way.
This guide will show you how to modify your existing workflows to easily get a simulation running with multiple levels!
Getting Started
Utilizing multiple levels is easy! Taking any existing simulation settings file, modify the following parameters:
"Scenario": {
"Name": "DataCollection",
"Options": {
"MultiLevel": true,
"LevelNames": [
"Home1",
"Home2",
"Home3"
]
}
},
"Environment": {
"Name": "SWARMHome",
"StartingLevelName": "Home1",
"StreamVideo": true
},
Notice here that we have added a few different options to the Scenario section. You will need to always have the MultiLevel parameter and the LevelNames parameter.
For the section labelled Environment, ensure that you select a Level name that is in the Scenario Options section. For example, if you have the levels listed in the Scenario section as:
"LevelNames": [
"Home1",
"Home2",
"Home3"
]
you should choose only from that list when selecting the StartingLevelName.
Execution
The SWARM Developer system manages the entire state of the simulation on any particular level. Currently, if executing the Data Collection scenario, the trajectory of the agent will be tracked and a new level is loaded when the agent has successfully completed the expected trajectory.
Each Scenario as defined in Supported Scenarios has an objective function that can be used to trigger the change to multiple levels.
Analysis and Data
If you selected to collect any data from the system, such as images or videos, then each image or video will be labelled by the level, along with the custom name that you provide. For Vehicle State information, the information is broken up by statements that are recorded when the levels are changed. The SWARM Analysis system automatically handles this case and splits the data appropriately.
To find out more about the data analysis capabilities, please visit SWARM Data Analysis