- Added 📥: BoundingBoxOriented Is a SensorCallback which projects all visible bounding boxes into the sensor frame of reference given a LiDAR sensor. This callback is currently only available for LiDAR subclasses. The class can be found in the simulation/sensors_callbacks.py file. The class works with the PointCloudVisualizer class and displays the bounding boxes as Open3D OrientedBoundingBoxes in the point cloud data.
- Added 📥: CarlaObject used as a wrapper for carla.CityObjectLabel. The class can be found in the utils/carla_objects.py file.
- Added 📥: CarlaObjects a container for all CarlaObject instances. The class can be found in the utils/carla_objects.py file. It is used to store all CarlaObject instances in a single object for easy access. When creating a custom type of CarlaObject it is important to add the new type to the CarlaObjects class.
- Added 📥: CarlaObjectCategory contains a group of CarlaObject instances. The class can be found in the utils/carla_objects.py file. There are some predefined categories available.
- Added 📥: CarlaObjectCategories a container for all CarlaObjectCategory instances. The class can be found in the utils/carla_objects.py file. It is used to store all CarlaObjectCategory instances in a single object for easy access.
- Added 📥: Object coloring during visualization. Each color can be found and edited in the utils/constants.py file. The colors are stored in each CarlaObject instance. Easily accessed through the dataclass CarlaObjects.
- Added 📥: CarlaActor which is a wrapper for carla.Actor. The class can be found in the simulation.environment.carla_actor.py file. An instance can be created from either a carla.Actor or a carla.EnvironmentObject. This solves an issue where the carla.Actor and carla.EnvironmentObject does not contain similar attributes. This is to be able to keep track of all objects in the simulation.
- Added 📥: StationaryActor dummy class to wrapp the carla.EnvironmentObject class to give it similar attributes as the carla.Actor class. The class can be found in the simulation.environment.carla_actor.py file. This is to be able to keep track of all objects in the simulation.
- Added 📥: CarlaTrafficSpawner which can spawn an arbitrary number of vehicles and pedestrians in the simulation. The class can be found in the simulation/environment/traffic.py file.
- Added 📥: CarlaActorManager which can read all spawned or default objects and create a CarlaActor for each instance with the correct CarlaObject label.
- Removed 🗑️: BoundingBoxOpen3D has been removed and replaced by BoundingBoxOriented.
- Performance 🚀: BoundingBoxProjector and it's subclasses now initalizes all bounding boxes based on a list of CarlaActors. It is usefull an easy to retrieve this list using a CarlaActorManager.
- Fixed 🔧: Fixed a bug in multi_3D.py where the frames would lag behind. This was due to a world.tick() before the rendering of the current frame. Changing order reduced the amount of frames droped due to latency.
- Added 📥: BoundingBoxOpen3D a SensorCallback which projects all visible bounding boxes onto the point cloud data. This callback is currently only available for LiDAR subclasses. The class can be found in the simulation/sensors_callbacks.py file. The class works with the PointCloudVisualizer class and displays the bounding boxes as points in the point cloud data.
- Added 📥: LiDAR_Open3D SensorCallback to all 3D LiDAR sensors to convert the scene of the LiDAR object to an Open3D point cloud. The class can be found in the simulation/lidars.py file.
- Added 📥: Properties to the Sensor class;
- is_synched - Boolean value indicating if the sensor is visualized or not.
- available_options - List of all available options for the sensor.
- default_options - List of all default options for the sensor.
- sensor_range - The range of the sensor.
- frame_id - The frame id of the sensor.
- Added 📥: classmethod SensorInformation.get_sensor_options to get the available options for a sensor. The method can be found in the utils/constants.py file.
- Changed 🔄: Sensor structure to utilize a python Queue to keep the sensors and visualizers thread safe. The sensor now adds the data to the queue and the visualizer retrieves the data from the queue. This allows for multiple visualizers to be used at once. The DisplayManager class can be found in the visualization/display_manager.py file.
- Fixed 🔧: Fixed a bug where frames would create backlog in the visualization face creating small lag spikes and unsynchronized frames between sensors and world data. Now skips frames if the visualization is not up to date with the world data.
- Added 📥: Added PointCloudVisualizer class for visualizining point cloud data in 3D. The class can be found in the visualization/point_cloud_visualizer.py file.
- Added 📥: LiDAR sensor now includes the intensity of the point cloud data in its visualization, both 2D and 3D. The intensity is visualized as a color gradient from blue to red. The intensity is also included in the point cloud data. The intensity is stored in the 4th channel of the point cloud data.
- Added 📥: SensorCallback class for post processing of sensor data before rendering. The class can be found in the simulation/sensors.py file.
- Added 📥: BoundingBoxProjector a SensorCallback which projects all visible bounding boxes onto the sensor image. This callback is currently only available for Camera subclasses. The class can be found in the simulation/sensors_callbacks.py file.
- Added 📥: docs/docstring_style.py showcasing the standard docstring style for the project.
- Changed 🔄: The Sensor class now takes a list of SensorCallback objects as an argument. The callbacks are called in the order they are given. The Sensor class can be found in the simulation/sensors.py file.
- Refactoring 📦: Altered the relation between the Camera class and its subclasses. The subclasses now initialize by sending a blueprint and data format parameter to the Camera class. The Camera class then creates the sensor and assigns the data format. The Camera class can be found in the simulation/sensors.py file.
- Style 🎨: Added a consitent class and function documenation style for the entire project. The style can be found in the utils/docstring_style.py file. The style is based on the Numpy docstring style.
- Performance 🚀: The DisplayManager renders now operates in a seperate thread. This inceases efficiency and allows for several visualizers at once. This also allows for the DisplayManager to be used in a non-blocking manner.
- Fixed 🔧: Fixed a bug where the DisplayManager would render the raw image data instead of the post processed image data. This bug was caused by the DisplayManager not waiting for the sensor to finish processing the data before rendering it and simply retieving the scene attribute of the camera sensor. This bug is now fixed by only assigning the scene attribute of the camera sensor when the sensor has finished processing the data.
- Fixed 🔧: Fixed a bug where the DisplayManager would not render the bounding boxes of the BoundingBoxProjector callback. This bug was caused by the DisplayManager not waiting for the sensor to finish processing the data before rendering it and simply retieving the scene attribute of the camera sensor. This bug is now fixed by only assigning the scene attribute of the camera sensor when the sensor has finished processing the data.
- Maintenance 🧹: Cleaned up the codebase and removed unused variables and imports. Also cleaned the utils/constants.py file and removed unused constants.
- Added 📥: Added a new class called SensorOptions which contains all the available options for a sensor. The class can be found in the utils/constants.py file.
- Changed 🔄: Altered the structure of the constants regarding sensor information. The constants can be found in the utils/constants.py file. Each sensor has its own class with the constants for that sensor. The constants now contain an 'OPTIONS' attribute which contains all available CARLA blueprint options and there default value. See the SensorOptions class in utils/constants.py.
- Performance 🚀: Optimized the code for the DisplayManager and Sensor classes. The DisplayManager class now only updates the display when the sensor has new data. The Sensor class now only updates the sensor when the sensor has new data. This results in a significant performance boost.
- Added 📥: Sensor wrappers for RGB Camera, Depth Camera, Semantic Segmentation Camera, LiDAR, Semantic LiDAR. The classes can be found in the simulation/sensors.py
- Added 📥: Modified visualization of all the sensors to be more modular. The visualizer was based on the example visualizer found in the CARLA repository
- Added 📥: Added traffic generation using example script from CARLA repository found here
| Category | Example description |
|---|---|
| Added 📥 | Added new feature X |
| Changed 🔄 | Updated function Y to improve performance |
| Removed 🗑️ | Removed deprecated API Z |
| Fixed 🔧 | Fixed bug causing application crash |
| Security 🔒 | Implemented enhanced encryption for user data |
| Deprecated 📛 | Marked method A as deprecated, use method B instead |
| Breaking Changes 🚨 | Renamed class C, update references accordingly |
| Documentation 📚 | Updated API documentation for better clarity |
| Maintenance 🧹 | Cleaned up codebase, removed unused variables |
| Performance 🚀 | Optimized database queries for faster response time |
| Refactoring 📦 | Extracted reusable components from module D |
| Style 🎨 | Applied consistent coding style across the project |
| Tests 🧪 | Added unit tests for module E |
| Other 📦 | Miscellaneous updates and improvements |