URDF and Sensors

This section describes the placement of sensors in EgoVehicle on the example of a Lexus RX450h 2015 Sample Sensor prefab.

URDF (Unified Robot Description Format) is equivalent to the simplified URDF format used in ROS2. This format allows to define the positions of all sensors of the vehicle in relation to its main parent prefab coordinate system.

URDF is added directly to the main parent of the prefab and there are no transforms between these objects. It is built using multiple GameObjects as children appropriately transformed with relation to its parent.

The transforms in the URDF object are defined using the data from the sensor kit documentation used in the vehicle. Such data can be obtained from sensor kit packages for Autoware, for example: awsim_sensor_kit_launch - it is used in the AWSIM compatible version of Autoware. This package contains a description of transforms between coordinate systems (frames) in the form of *.yaml files: sensors_calibration and sensor_kit_calibration.

In the first file, the transform of the sensor kit frame (sensor_kit_base_link) relative to the local vehicle frame (base_link) is defined. In Unity, this transform is defined in the object Sensor Kit. While the second file contains a definition of the transformations of all sensors with respect to the sensor kit - they are described in the Sensor Kit subsections.

Transformations

Please note that the transformation Objects are intended to be a direct reflection of frames existing in ROS2. All frame Objects are defined as children of base_link and consist of nothing but a transformation - analogical to the one present in ROS2 (keep in mind the coordinate system conversion). The sensor Objects are added to the transformation Object with no transformation of their own.

Coordinate system conventions

Unity uses a left-handed convention for its coordinate system, while the ROS2 uses a right-handed convention. For this reason, you should remember to perform conversions to get the correct transforms.

Base Link

Base Link (frame named base_link) is the formalized local coordinate system in URDF. All sensors that publish data specified in some frame present in Autoware are defined in relation to base_link. It is a standard practice in ROS, that base_link is a parent transformation of the whole robot and all robot parts are defined in some relation to the base_link.

If any device publishes data in the base_link frame - it is added as a direct child, with no additional transformation intermediate Object (PoseSensor is an example). However, if this device has its own frame, it is added as a child to its frame Object - which provides an additional transformation. The final transformation can consist of many intermediate transformation Objects. The frame Objects are added to the base_link (GnssSensor and its parent gnss_link are an example).

Sensor Kit

Sensor Kit (frame named sensor_kit_base_link) is a set of objects that consists of all simulated sensors that are physically present in an autonomous vehicle and have their own coordinate system (frame). This set of sensors has its own frame sensor_kit_base_link that is relative to the base_link.

In the Lexus RX450h 2015 Sample Sensor prefab, it is added to the base_link GameObject with an appropriately defined transformation. It acts as an intermediate frame GameObject. Sensor Kit is located on the top of the vehicle, so it is significantly shifted about the Oy and Oz axes. Sensors can be defined directly in this Object (VelodyneVLP16 is an example), or have their own transformation Object added on top of the sensor_kit_base_link (like GnssSensor mentioned in the base_link section).

The sensors described in this subsection are defined in relation to the sensor_kit_base_link frame.

LiDARs

LidarSensor is the component that simulates the LiDAR (Light Detection and Ranging) sensor. The LiDARs mounted on the top of autonomous vehicles are primarily used to scan the environment for localization in space and for detection and identification of obstacles. LiDARs placed on the left and right sides of the vehicle are mainly used to monitor the traffic lane and detect vehicles moving in adjacent lanes. A detailed description of this sensor is available in this section.

Lexus RX450h 2015 Sample Sensor prefab has one VelodyneVLP16 prefab sensor configured on the top of the vehicle, mainly used for location in space, but also for object recognition. Since the top LiDAR publishes data directly in the sensor_kit_base_link frame, the prefab is added directly to it - there is no transform. The other two remaining LiDARs are defined, but disabled - they do not provide data from space (but you can enable them!).

Top

Left - disabled

Right - disabled

IMU

IMUSensor is a component that simulates an IMU (Inertial Measurement Unit) sensor. It measures acceleration and angular velocity of the EgoVehicle. A detailed description of this sensor is available in this section.

Lexus RX450h 2015 Sample Sensor has one such sensor located on the top of the vehicle. It is added to an Object tamagawa/imu_link that matches its frame_id and contains its transform with respect to sensor_kit_base_link. This transformation has no transition, but only rotation around the Oy and Oz axes. The transform is defined in such a way that its axis Oy points downwards - in accordance with the gravity vector.

GNSS

GnssSensor is a component which simulates the position of vehicle computed by the Global Navigation Satellite. A detailed description of this sensor is available in this section.

Lexus RX450h 2015 Sample Sensor prefab has one such sensor located on top of the vehicle. It is added to an Object gnss_link that matches its frame_id and contains its transform with respect to sensor_kit_base_link. The frame is slightly moved back along the Oy and Oz axes.

Camera

CameraSensor is a component that simulates an RGB camera. Autonomous vehicles can be equipped with many cameras used for various purposes. A detailed description of this sensor is available in this section.

Lexus RX450h 2015 Sample Sensor prefab has one camera, positioned on top of the vehicle in such a way that the cameras field of view provides an image including traffic lights - the status of which must be recognized by Autoware. It is added to an Object traffic_light_left_camera/camera_link that matches its frame_id and contains its transform with respect to sensor_kit_base_link.

Pose

PoseSensor is a component which provides access to the current position and rotation of the EgoVehicle - is added as a ground truth.

The position and orientation of EgoVehicle is defined as the position of the frame base_link in the global frame, so this Object is added directly as its child without a transform.

VehicleSensor

VehicleStatusSensor is a component that is designed to aggregate information about the current state of the EgoVehicle, such as the active control mode, vehicle speed, steering of its wheels, or turn signal status. A detailed description of this sensor is available in this section.

This Object is not strictly related to any frame, however, it is assumed as a sensor, therefore it is added to the URDF.