Parameters

This section describes how to configure the topics that scenario_simulator_v2 publishes to Autoware.

Overview

The topics that scenario_simulator_v2 publishes to Autoware are configurable from the ROS 2 parameter file given to the launch argument parameter_file_path of scenario_test_runner. The default value of parameter_file_path is the path to a sample parameter file.

All parameters that can be specified and their default values are shown in the sample parameter file. In practice, it is not necessary to specify all parameters except for some that are mandatory. In that case, the simulator will behave as if similar default values had been specified.

There are currently two ways to configure some topics: an old way and a new way described on this page. The new way is backward compatible and is the recommended way. If you want to know how to use the old way, see this page.

/perception/object_recognition/detection/objects

version

An int type value in YYYYMMDD format, mandatory. Suffix of scenario_test_runner launch argument architecture_type, used to maintain backward compatibility of the simulator when changing the Autoware interface.

seed

A positive int type value, default 0. The seed value for the random number generator. If 0 is specified, a random seed value will be generated for each run.

override_legacy_configuration

A boolean type value, default false. Some of the parameters described below can be configured in either the old or new way. This parameter is used to determine which value to use. That is, as long as this parameter is false, some of the following parameters will be ignored and the values set by the old method will be used. If you want to configure the new way, set it to true. For backward compatibility, the default value of this parameter is false.

delay

A positive double type value, default 0.0. The unit is seconds. It is an error if the value is negative. Delays the publication of the topic by the specified number of seconds. This parameter is used only if override_legacy_configuration is true. If it is false, the value of detectedObjectPublishingDelay in ObjectController.Properties in the scenario file is used.

range

A positive double type value, default 300.0. The unit is meters. The sensor detection range. This parameter is used only if override_legacy_configuration is true. If it is false, the value of detectionSensorRange in ObjectController.Properties in the scenario file is used.

occlusionless

A boolean type value, default false. The message is a simulated object recognition result based on a pointcloud. Pointclouds are usually sensed by LiDAR, and scenario_simulator_v2 assumes this and simulates it, including LiDAR occlusion. If this parameter is true, object recognition is simulated as if there is no occlusion. In other words, it produces recognition results as if objects behind the object are also visible (even though they are in shadow and invisible in normal LiDAR). This parameter is used only if override_legacy_configuration is true. If it is false, the value of isClairvoyant in ObjectController.Properties in the scenario file is used.

noise.model.version

A positive int type value, default 1. If a non-existent version is specified, it is an error. This parameter specifies the version of the noise model to be used. Currently, the following two noise models are implemented: - version: 1 - Simple noise model with position randomization - version: 2 - Elliptically approximated model of noise variation with distance from the ego entity

The parameters specific to the models are placed under noise.v1. and noise.v2, respectively.

noise.v1.position.standard_deviation

A positive double type value, default 0.0. Standard deviation used for randomization of the position of the vehicle in the message. This parameter is used only if the value of noise.model.version is 1.

noise.v1.missing_probability

A double type value between 0.0 and 1.0, default 0.0. Based on the probability specified by the value of this parameter, random vehicle data is removed from the message. This parameter is used only if the value of noise.model.version is 1.

noise.v2.ellipse_y_radii

Array of positive double type values, default [10.0, 20.0, 40.0, 60.0, 80.0, 120.0, 150.0, 180.0, 1000.0]. Units are in meters. The size of the array is arbitrary, but must be the same size as the array described later. This parameter is used only if the value of noise.model.version is 2.

noise.v2.distance.autocorrelation_coefficient.amplitude

A positive double type value, default 0.0. The parameter of the autocorrelation coefficient used in the generation of distance noise. The autocorrelation coefficient $\phi$ is calculated by the following equation: $<span class="arithmatex"><span class="MathJax_Preview">\phi(\varDelta t) = \mathtt{amplitude} * \exp(-\mathtt{decay} * \varDelta t) + \mathtt{offset}</span><script type="math/tex">\phi(\varDelta t) = \mathtt{amplitude} * \exp(-\mathtt{decay} * \varDelta t) + \mathtt{offset}$ The noise models the time series as AR(1) model. The autocorrelation coefficient $\phi$ is used in the model to calculate the position noise $X_\mathrm{distance}$: as follows: $<span class="arithmatex"><span class="MathJax_Preview">X_\mathrm{distance}(t) = \mathtt{mean} + \phi * (X_\mathrm{distance}(t-1) - \mathtt{mean}) + \mathcal{N}(0, 1 - \phi^2) * \mathtt{standard\_deviation}</span><script type="math/tex">X_\mathrm{distance}(t) = \mathtt{mean} + \phi * (X_\mathrm{distance}(t-1) - \mathtt{mean}) + \mathcal{N}(0, 1 - \phi^2) * \mathtt{standard\_deviation}$ This parameter is used only if the value of noise.model.version is 2.

noise.v2.distance.autocorrelation_coefficient.decay

A positive double type value, default 0.0. The parameter of the autocorrelation coefficient used in the generation of distance noise. The autocorrelation coefficient $\phi$ is calculated by the following equation: $<span class="arithmatex"><span class="MathJax_Preview">\phi(\varDelta t) = \mathtt{amplitude} * \exp(-\mathtt{decay} * \varDelta t) + \mathtt{offset}</span><script type="math/tex">\phi(\varDelta t) = \mathtt{amplitude} * \exp(-\mathtt{decay} * \varDelta t) + \mathtt{offset}$ The noise models the time series as AR(1) model. The autocorrelation coefficient $\phi$ is used in the model to calculate the position noise $X_\mathrm{distance}$: as follows: $<span class="arithmatex"><span class="MathJax_Preview">X_\mathrm{distance}(t) = \mathtt{mean} + \phi * (X_\mathrm{distance}(t - \varDelta t) - \mathtt{mean}) + \mathcal{N}(0, 1 - \phi^2) * \mathtt{standard\_deviation}</span><script type="math/tex">X_\mathrm{distance}(t) = \mathtt{mean} + \phi * (X_\mathrm{distance}(t - \varDelta t) - \mathtt{mean}) + \mathcal{N}(0, 1 - \phi^2) * \mathtt{standard\_deviation}$ This parameter is used only if the value of noise.model.version is 2.

noise.v2.distance.autocorrelation_coefficient.offset

A positive double type value, default 0.0. The parameter of the autocorrelation coefficient used in the generation of distance noise. The autocorrelation coefficient $\phi$ is calculated by the following equation: $<span class="arithmatex"><span class="MathJax_Preview">\phi(\varDelta t) = \mathtt{amplitude} * \exp(-\mathtt{decay} * \varDelta t) + \mathtt{offset}</span><script type="math/tex">\phi(\varDelta t) = \mathtt{amplitude} * \exp(-\mathtt{decay} * \varDelta t) + \mathtt{offset}$ The noise models the time series as AR(1) model. The autocorrelation coefficient $\phi$ is used in the model to calculate the position noise $X_\mathrm{distance}$: as follows: $<span class="arithmatex"><span class="MathJax_Preview">X_\mathrm{distance}(t) = \mathtt{mean} + \phi * (X_\mathrm{distance}(t - \varDelta t) - \mathtt{mean}) + \mathcal{N}(0, 1 - \phi^2) * \mathtt{standard\_deviation}</span><script type="math/tex">X_\mathrm{distance}(t) = \mathtt{mean} + \phi * (X_\mathrm{distance}(t - \varDelta t) - \mathtt{mean}) + \mathcal{N}(0, 1 - \phi^2) * \mathtt{standard\_deviation}$ This parameter is used only if the value of noise.model.version is 2.

noise.v2.distance.mean.ellipse_normalized_x_radius

A positive double type value, default 0.0. The noise models the space as an elliptical model. This parameter is the ratio of the radius of the x-axis to the radius of the y-axis of that ellipse. The coordinate system is a right-handed local coordinate system, where the x-axis is the longitudinal direction of the ego entity and the y-axis is its lateral direction. The value of this parameter is used to calculate the distance $d$ between the ego entity and the other vehicle using the following equation: $<span class="arithmatex"><span class="MathJax_Preview">d = \sqrt[2]{(\varDelta x / \mathtt{ellipse\_normalized\_x\_radius})^2 + \varDelta y^2}</span><script type="math/tex">d = \sqrt[2]{(\varDelta x / \mathtt{ellipse\_normalized\_x\_radius})^2 + \varDelta y^2}$ This parameter is used only if the value of noise.model.version is 2.

noise.v2.distance.mean.values

Array of positive double type values, default [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]. Each element of the array is a mean of normal distribution. The first element with a value greater than $d$ is searched from ellipse_y_radii and the elements with the same index are referenced from values. Therefore, the array size of this parameter must be the same as ellipse_y_radii. Otherwise, it is an error. This parameter is used only if the value of noise.model.version is 2.

noise.v2.distance.standard_deviation.ellipse_normalized_x_radius

A positive double type value, default 0.0. The noise models the space as an elliptical model. This parameter is the ratio of the radius of the x-axis to the radius of the y-axis of that ellipse. The coordinate system is a right-handed local coordinate system, where the x-axis is the longitudinal direction of the ego entity and the y-axis is its lateral direction. The value of this parameter is used to calculate the distance $d$ between the ego entity and the other vehicle using the following equation: $<span class="arithmatex"><span class="MathJax_Preview">d = \sqrt[2]{(\varDelta x / \mathtt{ellipse\_normalized\_x\_radius})^2 + \varDelta y^2}</span><script type="math/tex">d = \sqrt[2]{(\varDelta x / \mathtt{ellipse\_normalized\_x\_radius})^2 + \varDelta y^2}$ This parameter is used only if the value of noise.model.version is 2.

noise.v2.distance.standard_deviation.values

Array of positive double type values, default [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]. Each element of the array is a standard deviation of normal distribution. The first element with a value greater than $d$ is searched from ellipse_y_radii and the elements with the same index are referenced from values. Therefore, the array size of this parameter must be the same as ellipse_y_radii. Otherwise, it is an error. This parameter is used only if the value of noise.model.version is 2.

noise.v2.yaw.autocorrelation_coefficient.amplitude

A positive double type value, default 0.0. The parameter of the autocorrelation coefficient used in the generation of yaw noise. The autocorrelation coefficient $\phi$ is calculated by the following equation: $$\phi(\varDelta t) = \mathtt{amplitude} * \exp(-\mathtt{decay} * \varDelta t) + \mathtt{offset}$$ The noise models the time series as AR(1) model. The autocorrelation coefficient $\phi$ is used in the model to calculate the yaw noise $X_\mathrm{yaw}$: as follows: $$X_\mathrm{yaw}(t) = \mathtt{mean} + \phi * (X_\mathrm{yaw}(t - \varDelta t) - \mathtt{mean}) + \mathcal{N}(0, 1 - \phi^2) * \mathtt{standard\_deviation}$$ This parameter is used only if the value of noise.model.version is 2.

noise.v2.yaw.autocorrelation_coefficient.decay

A positive double type value, default 0.0. The parameter of the autocorrelation coefficient used in the generation of yaw noise. The autocorrelation coefficient $\phi$ is calculated by the following equation: $$\phi(\varDelta t) = \mathtt{amplitude} * \exp(-\mathtt{decay} * \varDelta t) + \mathtt{offset}$$ The noise models the time series as AR(1) model. The autocorrelation coefficient $\phi$ is used in the model to calculate the yaw noise $X_\mathrm{yaw}$: as follows: $$X_\mathrm{yaw}(t) = \mathtt{mean} + \phi * (X_\mathrm{yaw}(t - \varDelta t) - \mathtt{mean}) + \mathcal{N}(0, 1 - \phi^2) * \mathtt{standard\_deviation}$$ This parameter is used only if the value of noise.model.version is 2.

noise.v2.yaw.autocorrelation_coefficient.offset

A positive double type value, default 0.0. The parameter of the autocorrelation coefficient used in the generation of yaw noise. The autocorrelation coefficient $\phi$ is calculated by the following equation: $$\phi(\varDelta t) = \mathtt{amplitude} * \exp(-\mathtt{decay} * \varDelta t) + \mathtt{offset}$$ The noise models the time series as AR(1) model. The autocorrelation coefficient $\phi$ is used in the model to calculate the yaw noise $X_\mathrm{yaw}$: as follows: $$X_\mathrm{yaw}(t) = \mathtt{mean} + \phi * (X_\mathrm{yaw}(t - \varDelta t) - \mathtt{mean}) + \mathcal{N}(0, 1 - \phi^2) * \mathtt{standard\_deviation}$$ This parameter is used only if the value of noise.model.version is 2.

noise.v2.yaw.mean.ellipse_normalized_x_radius

A positive double type value, default 0.0. The noise models the space as an elliptical model. This parameter is the ratio of the radius of the x-axis to the radius of the y-axis of that ellipse. The coordinate system is a right-handed local coordinate system, where the x-axis is the longitudinal direction of the ego entity and the y-axis is its lateral direction. The value of this parameter is used to calculate the distance $d$ between the ego entity and the other vehicle using the following equation: $<span class="arithmatex"><span class="MathJax_Preview">d = \sqrt[2]{(\varDelta x / \mathtt{ellipse\_normalized\_x\_radius})^2 + \varDelta y^2}</span><script type="math/tex">d = \sqrt[2]{(\varDelta x / \mathtt{ellipse\_normalized\_x\_radius})^2 + \varDelta y^2}$ This parameter is used only if the value of noise.model.version is 2.

noise.v2.yaw.mean.values

Array of positive double type values, default [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]. Each element of the array is a mean of normal distribution. The first element with a value greater than $d$ is searched from ellipse_y_radii and the elements with the same index are referenced from values. Therefore, the array size of this parameter must be the same as ellipse_y_radii. Otherwise, it is an error. This parameter is used only if the value of noise.model.version is 2.

noise.v2.yaw.standard_deviation.ellipse_normalized_x_radius

A positive double type value, default 0.0. The noise models the space as an elliptical model. This parameter is the ratio of the radius of the x-axis to the radius of the y-axis of that ellipse. The coordinate system is a right-handed local coordinate system, where the x-axis is the longitudinal direction of the ego entity and the y-axis is its lateral direction. The value of this parameter is used to calculate the distance $d$ between the ego entity and the other vehicle using the following equation: $<span class="arithmatex"><span class="MathJax_Preview">d = \sqrt[2]{(\varDelta x / \mathtt{ellipse\_normalized\_x\_radius})^2 + \varDelta y^2}</span><script type="math/tex">d = \sqrt[2]{(\varDelta x / \mathtt{ellipse\_normalized\_x\_radius})^2 + \varDelta y^2}$. This parameter is used only if the value of noise.model.version is 2.

noise.v2.yaw.standard_deviation.values

Array of positive double type values, default [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]. Each element of the array is a standard deviation of normal distribution. The first element with a value greater than $d$ is searched from ellipse_y_radii and the elements with the same index are referenced from values. Therefore, the array size of this parameter must be the same as ellipse_y_radii. Otherwise, it is an error. This parameter is used only if the value of noise.model.version is 2.

noise.v2.yaw_flip.autocorrelation_coefficient.amplitude

A positive double type value, default 0.0. The parameter of the autocorrelation coefficient used in the generation of yaw-flip noise. The autocorrelation coefficient $\phi$ is calculated by the following equation: $<span class="arithmatex"><span class="MathJax_Preview">\phi(\varDelta t) = \mathtt{amplitude} * \exp(-\mathtt{decay} * \varDelta t) + \mathtt{offset}</span><script type="math/tex">\phi(\varDelta t) = \mathtt{amplitude} * \exp(-\mathtt{decay} * \varDelta t) + \mathtt{offset}$ The noise models the time series as Markov process. The autocorrelation coefficient $\phi$ is used in the model to calculate the yaw-flip noise with following transition matrix: $$\begin{bmatrix} p_{0,0} & p_{0,1} \\ p_{1,0} & p_{1,1} \end{bmatrix} = \begin{bmatrix} \pi_0 + \phi \pi_1 && \pi_1 (1 - \phi) \\ \pi_0 (1 - \phi) && \pi_1 - \phi \pi_0 \end{bmatrix}$$ This parameter is used only if the value of noise.model.version is 2.

noise.v2.yaw_flip.autocorrelation_coefficient.decay

A positive double type value, default 0.0. The parameter of the autocorrelation coefficient used in the generation of yaw-flip noise. The autocorrelation coefficient $\phi$ is calculated by the following equation: $<span class="arithmatex"><span class="MathJax_Preview">\phi(\varDelta t) = \mathtt{amplitude} * \exp(-\mathtt{decay} * \varDelta t) + \mathtt{offset}</span><script type="math/tex">\phi(\varDelta t) = \mathtt{amplitude} * \exp(-\mathtt{decay} * \varDelta t) + \mathtt{offset}$ The noise models the time series as Markov process. The autocorrelation coefficient $\phi$ is used in the model to calculate the yaw-flip noise with following transition matrix: $$\begin{bmatrix} p_{0,0} & p_{0,1} \\ p_{1,0} & p_{1,1} \end{bmatrix} = \begin{bmatrix} \pi_0 + \phi \pi_1 && \pi_1 (1 - \phi) \\ \pi_0 (1 - \phi) && \pi_1 - \phi \pi_0 \end{bmatrix}$$ This parameter is used only if the value of noise.model.version is 2.

noise.v2.yaw_flip.autocorrelation_coefficient.offset

A positive double type value, default 0.0. The parameter of the autocorrelation coefficient used in the generation of yaw-flip noise. The autocorrelation coefficient $\phi$ is calculated by the following equation: $<span class="arithmatex"><span class="MathJax_Preview">\phi(\varDelta t) = \mathtt{amplitude} * \exp(-\mathtt{decay} * \varDelta t) + \mathtt{offset}</span><script type="math/tex">\phi(\varDelta t) = \mathtt{amplitude} * \exp(-\mathtt{decay} * \varDelta t) + \mathtt{offset}$ The noise models the time series as Markov process. The autocorrelation coefficient $\phi$ is used in the model to calculate the yaw-flip noise with following transition matrix: $$\begin{bmatrix} p_{0,0} & p_{0,1} \\ p_{1,0} & p_{1,1} \end{bmatrix} = \begin{bmatrix} \pi_0 + \phi \pi_1 && \pi_1 (1 - \phi) \\ \pi_0 (1 - \phi) && \pi_1 - \phi \pi_0 \end{bmatrix}$$ This parameter is used only if the value of noise.model.version is 2.

noise.v2.yaw_flip.speed_threshold

A positive double type value, default 0.1. When the absolute speed of one other vehicle is less than the value of this parameter, it is determined whether yaw-flip occurs or not based on the rate described below. This parameter is used only if the value of noise.model.version is 2.

noise.v2.yaw_flip.rate

A positive double type value, default 0.0. Vehicles whose absolute speed is below the aforementioned speed_threshold will have yaw-flip noise applied with the probability of the value of this parameter. This parameter is used only if the value of noise.model.version is 2.

noise.v2.true_positive.autocorrelation_coefficient.amplitude

A positive double type value, default 0.0. The parameter of the autocorrelation coefficient used in the generation of random-mask noise. The autocorrelation coefficient $\phi$ is calculated by the following equation: $<span class="arithmatex"><span class="MathJax_Preview">\phi(\varDelta t) = \mathtt{amplitude} * \exp(-\mathtt{decay} * \varDelta t) + \mathtt{offset}</span><script type="math/tex">\phi(\varDelta t) = \mathtt{amplitude} * \exp(-\mathtt{decay} * \varDelta t) + \mathtt{offset}$ The noise models the time series as Markov process. The autocorrelation coefficient $\phi$ is used in the model to calculate the random-mask noise with following transition matrix: $$\begin{bmatrix} p_{0,0} & p_{0,1} \\ p_{1,0} & p_{1,1} \end{bmatrix} = \begin{bmatrix} \pi_0 + \phi \pi_1 && \pi_1 (1 - \phi) \\ \pi_0 (1 - \phi) && \pi_1 - \phi \pi_0 \end{bmatrix}$$ This parameter is used only if the value of noise.model.version is 2.

noise.v2.true_positive.autocorrelation_coefficient.decay

A positive double type value, default 0.0. The parameter of the autocorrelation coefficient used in the generation of random-mask noise. The autocorrelation coefficient $\phi$ is calculated by the following equation: $<span class="arithmatex"><span class="MathJax_Preview">\phi(\varDelta t) = \mathtt{amplitude} * \exp(-\mathtt{decay} * \varDelta t) + \mathtt{offset}</span><script type="math/tex">\phi(\varDelta t) = \mathtt{amplitude} * \exp(-\mathtt{decay} * \varDelta t) + \mathtt{offset}$ The noise models the time series as Markov process. The autocorrelation coefficient $\phi$ is used in the model to calculate the random-mask noise with following transition matrix: $$\begin{bmatrix} p_{0,0} & p_{0,1} \\ p_{1,0} & p_{1,1} \end{bmatrix} = \begin{bmatrix} \pi_0 + \phi \pi_1 && \pi_1 (1 - \phi) \\ \pi_0 (1 - \phi) && \pi_1 - \phi \pi_0 \end{bmatrix}$$ This parameter is used only if the value of noise.model.version is 2.

noise.v2.true_positive.autocorrelation_coefficient.offset

A positive double type value, default 0.0. The parameter of the autocorrelation coefficient used in the generation of random-mask noise. The autocorrelation coefficient $\phi$ is calculated by the following equation: $<span class="arithmatex"><span class="MathJax_Preview">\phi(\varDelta t) = \mathtt{amplitude} * \exp(-\mathtt{decay} * \varDelta t) + \mathtt{offset}</span><script type="math/tex">\phi(\varDelta t) = \mathtt{amplitude} * \exp(-\mathtt{decay} * \varDelta t) + \mathtt{offset}$ The noise models the time series as Markov process. The autocorrelation coefficient $\phi$ is used in the model to calculate the random-mask noise with following transition matrix: $$\begin{bmatrix} p_{0,0} & p_{0,1} \\ p_{1,0} & p_{1,1} \end{bmatrix} = \begin{bmatrix} \pi_0 + \phi \pi_1 && \pi_1 (1 - \phi) \\ \pi_0 (1 - \phi) && \pi_1 - \phi \pi_0 \end{bmatrix}$$ This parameter is used only if the value of noise.model.version is 2.

noise.v2.true_positive.rate.ellipse_normalized_x_radius

A positive double type value, default 0.0. The noise models the space as an elliptical model. This parameter is the ratio of the radius of the x-axis to the radius of the y-axis of that ellipse. The coordinate system is a right-handed local coordinate system, where the x-axis is the longitudinal direction of the ego entity and the y-axis is its lateral direction. The value of this parameter is used to calculate the distance $d$ between the ego entity and the other vehicle using the following equation: $<span class="arithmatex"><span class="MathJax_Preview">d = \sqrt[2]{(\varDelta x / \mathtt{ellipse\_normalized\_x\_radius})^2 + \varDelta y^2}</span><script type="math/tex">d = \sqrt[2]{(\varDelta x / \mathtt{ellipse\_normalized\_x\_radius})^2 + \varDelta y^2}$. This parameter is used only if the value of noise.model.version is 2.

noise.v2.true_positive.rate.values

Array of positive double type values, default [1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0]. Each element of this array is the probability that the value will be output correctly (true positive rate). The first element with a value greater than $d$ is searched from ellipse_y_radii and the elements with the same index are referenced from values. Therefore, the array size of this parameter must be the same as ellipse_y_radii. Otherwise, it is an error. This parameter is used only if the value of noise.model.version is 2.

/perception/object_recognition/ground_truth/objects

version

An int type value in YYYYMMDD format, mandatory. Suffix of scenario_test_runner launch argument architecture_type, used to maintain backward compatibility of the simulator when changing the Autoware interface.

override_legacy_configuration

A boolean type value, default false. Some of the parameters described below can be configured in either the old or new way. This parameter is used to determine which value to use. That is, as long as this parameter is false, some of the following parameters will be ignored and the values set by the old method will be used. If you want to configure the new way, set it to true. For backward compatibility, the default value of this parameter is false.

delay

A positive double type value, default 0.0. The unit is seconds. It is an error if the value is negative. Delays the publication of the topic by the specified number of seconds. This parameter is used only if override_legacy_configuration is true. If it is false, the value of detectedObjectGroundTruthPublishingDelay in ObjectController.Properties in the scenario file is used.