Add Sensors to RoadRunner Scenario Using MATLAB

Define sensor models in MATLAB®, and add them to vehicle actors in a RoadRunner Scenario. Then, obtain ground truth measurements from RoadRunner Scenario, process them into detections for visualization.

Set Up RoadRunner Scenario — MATLAB Interface

Configure your RoadRunner installation and project folder properties. Open the RoadRunner app.

rrInstallationPath = "C:\Program Files\RoadRunner R2024a\bin\win64";
rrProjectPath = "D:\RR\TestProjects";

s = settings;
s.roadrunner.application.InstallationFolder.PersonalValue = rrInstallationPath;
rrApp = roadrunner(rrProjectPath);

To open the scenario this example uses, you must add the TrajectoryCutIn-longRun.rrscenario file from the example folder to your RoadRunner project folder. Then, open the scenario.

copyfile("TrajectoryCutIn-longRun.rrscenario",fullfile(rrProjectPath,"Scenarios/"))
openScenario(rrApp,"TrajectoryCutIn-longRun")

Create a ScenarioSimulation object to connect MATLAB to the RoadRunner Scenario simulation and set the step size.

scenarioSim = createSimulation(rrApp);

Connection status: 1
Connected to RoadRunner Scenario server on localhost:60730, with client id {761e01bc-376c-4b4b-8ffa-aa1490d7438d}

stepSize = 0.1;
set(scenarioSim,"StepSize",stepSize);

Create a SensorSimulation object to control the sensor configuration for the RoadRunner Scenario simulation.

sensorSim = get(scenarioSim,"SensorSimulation");

To use the GPU on your device for supporting hardware accelerated raytracing, specify the UseGPU property of the SensorSimulation object to on. This enables supported sensors like lidars to use GPU for raytracing.

sensorSim.UseGPU = "on";

If you modify the RoadRunner scene after creating the SensorSimulation object, you must follow these steps for the changes to reflect correctly in MATLAB:

• Save and close RoadRunner application.

• Clear the ScenarioSimulation object from the MATLAB workspace.

• Reopen the RoadRunner scenario.

• Recreate the ScenarioSimulation object and the SensorSimulation object.

Configure Sensors and Add to RoadRunner Scenario

Configure sensor models for vision, radar and lidar sensors to add to the ego vehicle using visionDetectionGenerator, drivingRadarDataGenerator and lidarPointCloudGenerator objects. Specify unique IDs for each sensor.

visionSensor = visionDetectionGenerator(SensorIndex=1, ...
            SensorLocation=[2.4 0], MaxRange=50, ...
            DetectorOutput="Lanes and objects", ...
            UpdateInterval=stepSize);

radarSensor = drivingRadarDataGenerator(SensorIndex=2,...
    MountingLocation=[1.8 0 0.2], FieldOfView=[80 5],...
    AzimuthResolution=1,UpdateRate=1/stepSize);

lidarSensor = lidarPointCloudGenerator(SensorIndex=3,UpdateInterval=stepSize);

Add the sensor to the ego vehicle actor in the RoadRunner scenario. Specify the Actor ID property for the vehicle.

egoVehicleID = 1;
addSensors(sensorSim,{visionSensor,radarSensor,lidarSensor},egoVehicleID);

Simulate RoadRunner Scenario and Visualize Sensor Data

To visualize sensor data at each time-step of the simulation, add an observer to the RoadRunner scenario.

The helperSensorObserver system object implements the observer behavior. At the first timestep, the system object initializes the bird's-eye-plot for visualization, Then, at each time step, the system object:

1. Retrieves target poses in the sensor range using the targetPoses function.

2. Processes the target poses into detections using the sensor models.

3. Visualizes detections and ground truth lane boundaries using birdsEyePlot.

observer = addObserver(scenarioSim,"VisualizeSensorData","helperSensorObserver");

Start the scenario.

set(scenarioSim,"SimulationCommand","Start");

Helper Functions

helperSetupBEP function creates a bird's-eye-plot and configures all the plotters for visualization.

helperPlotLaneBoundaries function plots the lane boundaries on the birds'eye-plot.

helperSensorObserver system object implements the visualization of sensor data during the RoadRunner scenario simulation.

type("helperSensorObserver.m")

classdef helperSensorObserver < matlab.System

    properties(Access=private)
        currSimTime
        scenarioSimObj
        sensorSimObj
        visionSensor
        radarSensor
        lidarSensor
        visionDetPlotter
        radarDetPlotter
        pcPlotter
        lbGTPlotter
        lbDetPlotter
        bepAxes
    end

    methods
        % Constructor
        function obj = helperSensorObserver()
        end
    end

    methods(Access=protected)

        function interface = getInterfaceImpl(~)
            import matlab.system.interface.*;
            interface = ActorInterface;
        end
        
        % Get ScenarioSimulation and SensorSimulation objects, set up Bird's-Eye-Plot
        function setupImpl(obj)
            obj.scenarioSimObj = Simulink.ScenarioSimulation.find('ScenarioSimulation','SystemObject',obj);
            obj.sensorSimObj = obj.scenarioSimObj.get('SensorSimulation');
            
            obj.visionSensor = evalin("base","visionSensor");
            obj.radarSensor = evalin("base","radarSensor");
            obj.lidarSensor = evalin("base","lidarSensor");

            [obj.visionDetPlotter,obj.radarDetPlotter,obj.pcPlotter,obj.lbGTPlotter,obj.lbDetPlotter,obj.bepAxes] = helperSetupBEP(obj.visionSensor,obj.radarSensor);
            legend(obj.bepAxes,"show")

            obj.currSimTime = 0;
        end

        function releaseImpl(obj)
            obj.lidarSensor.release;
            obj.radarSensor.release;
            obj.visionSensor.release;
        end

        function stepImpl(obj)
            % Get current Simulation Time
            obj.currSimTime = obj.scenarioSimObj.get("SimulationTime");
            
            % Get ground truth target poses and lane boundaries from the sensor
            tgtPoses1 = targetPoses(obj.sensorSimObj,1);
            tgtPoses2 = targetPoses(obj.sensorSimObj,2);
            gTruthLbs = laneBoundaries(obj.sensorSimObj,1,OutputOption="EgoAdjacentLanes",inHostCoordinate=true);
            
            if ~isempty(gTruthLbs)
                
                % Get detections from vision and radar sensors
                [visionDets,numVisionDets,visionDetsValid,lbDets,numLbDets,lbDetsValid] = obj.visionSensor(tgtPoses1,gTruthLbs,obj.currSimTime);
                [radarDets,numRadarDets,radarDetsValid] = obj.radarSensor(tgtPoses2,obj.currSimTime);
        
                % Get point cloud from lidar sensor
                [ptCloud,ptCloudValid] = obj.lidarSensor();
        
                % Plot ground-truth and detected lane boundaries
                helperPlotLaneBoundaries(obj.lbGTPlotter,gTruthLbs)
               
                
                % Plot vision and radar detections
                if visionDetsValid
                    detPos = cellfun(@(d)d.Measurement(1:2),visionDets,UniformOutput=false);
                    detPos = vertcat(zeros(0,2),cell2mat(detPos')');
                    plotDetection(obj.visionDetPlotter,detPos)
                end
        
                if lbDetsValid
                    plotLaneBoundary(obj.lbDetPlotter,vertcat(lbDets.LaneBoundaries))
                end
        
                if radarDetsValid
                    detPos = cellfun(@(d)d.Measurement(1:2),radarDets,UniformOutput=false);
                    detPos = vertcat(zeros(0,2),cell2mat(detPos')');
                    plotDetection(obj.radarDetPlotter,detPos)
                end
        
                % Plot lidar point cloud
                if ptCloudValid
                    plotPointCloud(obj.pcPlotter,ptCloud);
                end
            end
        
        end  
    end

end