Gazebo Drivers

The gazebo drivers package stores all the scripts to fake the sensor data in gazebo.

scripts folder: Every sensor needs its own python script to work properly in the sim and if you are using the yolo faker to avoid gpu dependent tasks, you also need yaml camera files. These yamls are stored in the config folder and it also contains the actual calibration properties from the real world. These are used so that gazebo can properly distort the simulated image.

launch folder: Each launch file starts the python scripts of the appropriate sensor that the sub uses. These are dependent on sub and if new sensors are added, make sure to create the new python scripts along with adding them to the launch, so that gazebo can properly incorporate all the data the sub would experience.

yolo_faker

  • Starts ROS node YOLOFaker

This module contains the YOLOFaker node. It generates bounding boxes for objects in the sub simulator by taking in information about the camera and the objects bounding points. It listens for the ground truth poses from gazebo and broadcasts them as transforms in the world frame letting it transform the points in the object frame to the world frame to be projected in the camera frame. This lets it come up with a bounding box in the camera image mimicing what YOLO would would produce. This allows code development on statemachine or classical CV algorithms without the neural net being trained or without the use of a GPU.

To use this module you will need to provide a YAML configuration file specifying which cameras you want to use it on, their parameters, and which objects you want to have boxed. Objects can have multiple classes each with there own set of bounding boxes for bounding box generation. This allows for a single object to generate multiple hits like for example a start gate could have bounding points for each of its poles allowing them to be recognized individualy. For an example of the yaml file look at fakeyolo_default.yaml

To get the poses of the objects you want to track you will need to attach pose sensors to them in gazebo.

If you are using distortion in gazebo it is important to note that due to cropping of the barrel distortion your points will not be projected correctly unless you adjust the cameras field of view bigger than it actually is and then provide a camera matrix with the correct parameters which gazebo will not publish.

class YOLOObjectClass(name, color, points)

Detection class object

Holds object information for a bounding box class (i.e. the pole part of the startgate). This is for parts of objects that need to be classified rather than classifying the object as a whole.

class YOLOObject(frame_id, classes)

Holds object information required for bounding box generation

class Camera

Holds camera information required for bounding box generation

class YOLOFaker(object)

YOLO Faker Node

Takes in configuration yaml for camera, robot, and obstacle configurations and generates bounding boxes

Publishers

  • topic_name, type

  • bounding_boxes_topic, BoundingBoxes

  • debug_image_topic, Image

  • multiplexer_topic_debug, Image

Subscribers

  • topic_name, type

  • sub_pose_topic, Odometry

  • camera_info_topic, CameraInfo

  • image_topic, Image

  • multiplexer_topic, Image

  • odom_topic, Odometry

ActuatorService

class Actuator

Actuator service for simulator

  • Starts ROS node actuator_activate

Services

  • service_name, service_class

  • activateActuator, ActivateActuator

Service Proxies

  • service_name, service_class

  • /gazebo/spawn_urdf_model, SpawnModel

  • /gazebo/delete_model, DeleteModel

  • /gazebo/set_model_state, SetModelState

CameraInfoPublisher

class CameraInfoPublisher

Grabs ~camera_info_topic and ~frameid from parameter server and publishes them.

  • Starts ROS node CameraInfoPublisher

Service Proxies

  • service_name, service_class

  • ~camera_info_topic, CameraInfo

hydrophones_faker

class HydroFaker

Simulates hydrophones in Gazebo

should publish measurements periodically

if sub is facing w/in Psi degrees of direction of obstacle

  • meas noise ~N(0, sigma^2)

else

  • meas noise ~N(u, sigma^2), with some offset u ~ U(-45,45)

  • offset is preserved until sub moves position_delta meters away from that position

  • or sub turns yaw_delta meters away from where the offset was calculated

this simulates the random fluctuations of the pinger signal at each position in transdec

  • only one object at a time can be the source of the measurements

  • available objects on: cusub_common/hydrophones/debug_pinger_options

  • current object on: cusub_common/hydrophones/active

  • configured via service: cusub_common/hydrophones/switch_active_pinger

  • check whether the noise is biased: cusub_common/hydrophones/debug_is_biased

  • a pose msg for viewing in rviz is publish on: cusub_common/hydrophones/debug_measurements

  • Starts ROS node fakehydro

Services

  • service_name, service_class

  • hydrophones/switch_active_pinger, PingerSwitch

Publishers

  • topic_name, type

  • hydrophones/active, String

  • hydrophones/debug_is_biased, Bool

  • hydrophones/debug_pinger_options, PingerOptions

  • hydrophones/debug_measurements, PoseStamped

  • hydrophones/measurements, Hydrophones

Subscribers

  • topic_name, type

  • /leviathan/description/pose_gt, Odometry

  • fakehydro/objects/odom_topic, Odometry

depth_sensor

class DepthSensor

This node republishes of the depth sensor fluid pressure from gazebo into a depth pose for sensor fusion

  • Starts ROS node depth_sensor_gazebo

Publishers

  • topic_name, type

  • cusub_common/depth_odom, PoseWithCovarianceStamped

  • cusub_common/depth_map, PoseWithCovarianceStamped

Subscribers

  • topic_name, type

  • description/pressure, FluidPressure

dvl

class DVLRemap

Remaps and transforms the DVL data from gazebo to be like the real sub

  • Starts ROS node dvl_gazebo

Publishers

  • topic_name, type

  • cusub_common/dvl, TwistWithCovarianceStamped

Subscribers

  • topic_name, type

  • description/dvl_twist, TwistWithCovarianceStamped

motor_control

class MotorControl

This module takes motor pwm commands and mapps them to thrust commands for gazebo

  • Starts ROS node gazebo_motor_control

Publishers

  • topic_name, type

  • description/thrusters/<0-8>, FloatStamped

Subscribers

  • topic_name, type

  • cusub_common/motor_controllers/pololu_control/command, Float64MultiArray

z_odom_repub

class odom_callback

  • Starts ROS node new_odom_pub

Publishers

  • topic_name, type

  • sensor_fusion/odometry/filtered_z_flip, Odometry

Subscribers

  • topic_name, type

  • sensor_fusion/odometry/filtered, Odometry

thrust_odom

class ThrustOdom

Publishes thrust odometry to stabalize EKF

  • Starts ROS node thrust_odom_gazebo

Publishers

  • topic_name, type

  • cusub_common/thrust_odom, TwistWithCovarianceStamped

Subscribers

  • topic_name, type

  • cusub_common/motor_controllers/mux/drive/control_effort, Float64

  • cusub_common/motor_controllers/mux/strafe/control_effort, Float64

imu_repub

class IMURepub

  • Starts ROS node imu_repub

Publishers

  • topic_name, type

  • cusub_common/imu, Imu

Subscribers

  • topic_name, type

  • description/imu, Imu

  • mag_offset, Float64

Defined Message Types

  • ActivateActuator

    • int8 actuatorNumber

    • int32 activationTime

  • Hydrophones

    • std_msgs/Header header

    • float64 azimuth

  • PingerOptions

    • string[] options