Perception in Applications

The Application Simulator bundles a perception module for vehicle units. This module allows to emulate basic detection of other traffic entities using a field of view filter. To warrant fast simulation MOSAIC utilizes a spatial index, which allows for quick pre-selection of relevant entities.

Configuration

The perception module can be configured in the mosaic/scenarios/<scenario_name>/application/application_config.json. The most important configuration is the choice of a perception backend and its parameters. For every perceivable unit-type exists a defined index. The current configuration for the vehicle index happens via the parameter vehicleIndex, and for the traffic light index trafficLightIndex. Current index implementations are shown in the table below:

{
    "perceptionConfiguration": {
        "vehicleIndex": {
            "type": "VehicleGrid",
            "cellWidth": "5m",
            "cellHeight": "5m"
        },
        "trafficLightIndex": {
            "type": "TrafficLightMap"
        },
        "wallIndex": {
            "type": "WallTree"
        }
    }
}

For more information on choosing a backend for your scenario see here.

Application Configuration
In order to use the perception module from your application it has to be enabled first. Viewing angles can be defined between 0° and 360°, while the range has to be larger than 0. Configuration works analogously to the AdHoc- and Cell- Communication-Modules and is usually done at startup:

private final static VIEWING_ANGLE = 120; // [degree]
private final static VIEWING_RANGE = 100; // [meter]

@Override
public void onStartup() {
    // set up the configuration for the perception module
    SimplePerceptionConfiguration perceptionModuleConfiguration = 
        new SimplePerceptionConfiguration(VIEWING_ANGLE, VIEWING_RANGE);
    // enable the perception module using the defined configuration
    getOs().getPerceptionModule().enable(perceptionModuleConfiguration);  
}

Usage

Vehicles
To get a list of vehicles in perception range the getPerceivedVehicles()-method is called:

// get list of vehicles in perception range
List<VehicleObject> perceivedVehicles = getOs().getPerceptionModule().getPerceivedVehicles();
// log the list of perceived vehicle IDs
getLog().infoSimTime(this, "Perceived vehicles: {}",
        perceivedVehicles.stream().map(VehicleObject::getId).collect(Collectors.toList()));

The VehicleObject-class contains information about the perceived vehicles' position, speed, and heading, as well as its dimensions (length, width, height).

Traffic Lights
Retrieving all traffic lights in perception range is achieved using the getPerceivedTrafficLights():

// get list of traffic lights in perception range
List<TrafficLightObject> perceivedTrafficLights = getOs().getPerceptionModule().getPerceivedTrafficLights();
// log the list of perceived traffic light IDs
getLog().infoSimTime(this, "Perceived traffic lights: {}",
        perceivedTrafficLights.stream().map(TrafficLightObject::getId).collect(Collectors.toList()));

The TrafficLightObject-class contains information about the perceived traffic lights' position, state (i.e., green, red, …), and the incoming and outgoing lanes that are controlled by the individual signal.

Perception Modifiers

The perception module can be configured with different PerceptionModifiers, which can be used to emulate occlusion, false negatives, position areas, etc. MOSAIC already implements three modifiers, SimpleOcclusionModifier, WallOcclusionModifier, DistanceModifier and PositionErrorModifier.

Modifier	Description	Image
SimpleOcclusionModifier	Emulates occlusion in a simplified manner by comparing angles between perceived vehicles and requiring a minimum angle between all other perceived vehicles.
WallOcclusionModifier	Emulates occlusion of vehicles by buildings. Requires building information in the scenario database, which can be imported to the database using the --import-buildings option in scenario-convert.
DistanceModifier	Stochastic modifier that reduces perception probability with the distance to the ego vehicle.
PositionErrorModifier	Applies a gaussian error to lateral and longitudinal distances of perceived vehicles.

To configure modifiers they have to be passed to the PerceptionModuleConfiguration:

private void enablePerceptionModule() {
    // filter to emulate occlusion
    SimpleOcclusionModifier simpleOcclusionModifier = new SimpleOcclusionModifier(3, 5);
    // filter to emulate occlusion by buildings
    WallOcclusionModifier wallOcclusionModifier = new WallOcclusionModifier();
    // filter to reduce perception probability based on distance to ego vehicle
    DistanceModifier distanceModifier = new DistanceModifier(getRandom(), 0.0);
    // filter adding noise to longitudinal and lateral
    PositionErrorModifier positionErrorModifier = new PositionErrorModifier(getRandom());

    SimplePerceptionConfiguration perceptionModuleConfiguration = new SimplePerceptionConfiguration(
            VIEWING_ANGLE, VIEWING_RANGE,
            simpleOcclusionModifier, wallOcclusionModifier, distanceModifier, positionErrorModifier
    );
    getOs().getPerceptionModule().enable(perceptionModuleConfiguration);
}

All configured modifiers will be executed in order of configuration.