The supported compiler is clang, but that is not the default on Ubuntu.
Unreal Editor itself ships clang-10 for use on linux (along with the
libc++ that Unreal itself uses) so just use those in CARLA.
* Change handling of empty gbuffers.
Now, instead of sending an empty image, a black one with the size of the viewport is sent if the target gbuffer is unavailable.
* Add more GBufferID checks + improve empty gbuffer handling.
* Fix manual_control_gbuffer.py error on repeated listen_to_gbuffer.
* Expose is_listening and is_listening_gbuffer to the PythonAPI.
* Fix data race + autoremove unused gbuffers in manual_control_gbuffer.
* Update PythonAPI docs.
* Remove magic number.
* Switch from error to warning on invalid sensor type when requesting gbuffers.
* Add Misc/GBuffer
* Add EnqueueRenderSceneImmediateWithGBuffer.
* Modified ASceneCaptureSensor to allow gbuffer recording.
* Minor GBuffer progress.
* More GBuffer changes.
* Removed unnecessary files.
* Add FRHITexture* functions to FPixelReader + temporarily disabled non-gbuffer rendering in ASceneCaptureSensor.
* Add missing FPixelReader functions.
* Minor changes (Switching to Windows).
* Remove FRHITexture functions from FPixelReader, added the file ImageUtil.
* Remove FRHITexture functions from FPixelReader, added the file ImageUtil. (2)
* Added API to listen GBuffer data.
* Added gbuffer serializer classes
* Temporarily remove ViewRect hack.
* Add USceneCaptureComponent* derived classes.
* Disable USceneCaptureComponent*_CARLA and add initial FRHIGPUTextureReadBack-based code.
* Fix and re-enable custom SceneCaptureComponents.
* Fully switch to FRHIGPUTextureReadback.
* Remove unnecessary call to FlushRenderingCommands.
* Minor API changes.
* Add support for PF_DepthStencil in ImageUtil.
* More API progress...
* More API progress... (2)
* Removed testing code.
* Minor changes for testing.
* GBuffer API fixes.
* Improve GBuffer capture code.
* Fixed SceneDepth transfer issues and added SceneStencil, CustomDepth and CustomStencil to the GBuffer capture.
* Fix compilation error due to the usage of C++17 features.
* Removed major memory leak and added manual_control_gbuffer.py.
* Fixed a silly mistake.
* Minor changes to manual_control_gbuffer and SceneCaptureSensor.
* Fix compilation error on some versions of Ubuntu.
* Disable TAA when reading GBuffers to avoid jitter.
* Improve memory usage.
* Progress towards automatically detecting when a GBuffer stream is unused.
* Fix includes in SceneCaptureSensor + minor change in manual_control_gbuffer.py
* Progress on automatically detecting which GBuffers aren't needed.
* Remove unneeded __declspec.
* Revert ASensor changes + fix tutorial_gbuffer.py
* Update CHANGELOG.md
* Apply requested changes for the PR, add gitignore for the file OptionalModules.ini and add a GBufferTextureID enum to the Python API.
* Remove OptionalModules.ini.
* Fix indentation.
* Fix indentation (2).
* Fix indentation (3).
* Add documentation and more indentation fixes.
* Remove commented includes.
* Add missing line break.
* Fix memory leak + remove unneeded files.
* Add .uproject again, fix EngineAssociation.
* Remove unneeded ENQUEUE_RENDER_COMMAND.
* Fix manual_control_gbuffer.py.
* Add `stop_gbuffer` to the Python API.
* Minor fixes.
* Fix performance bug.
Previously, when a client requested a gbuffer that is unused it would stay open, even after stopping it explicitly. This commit fixes this issue.
* Fix indentation.
* Add missing braces, more indentation fixes and simplify some of the code.
* Update sensor.yml docs.
* Update docs.
* Remove unnecessary UE_Log + changed one verbosity level.
Co-authored-by: Axel <axellopez92@outlook.com>
Co-authored-by: Axel1092 <35765780+Axel1092@users.noreply.github.com>
* Added sparse map reprsentation
* Added pytorch for terramechanics modules
* Added saving loading particles in independent thread
* Added blank space at the EOF
* Added texture creation lambda
* Moving function to component
* Added input/output architecture for neural network
* Filling heightmap with particles data. Adding Logging to debug
* Updating Texture data at realtime
* Connected 3 stages, using MPC Position to Update and Texture
* Added square particle sampling. Added second model input/output scheme. Fixed start up crash.
* Added new flags and fixes
* Position to update not updating in material but yes in MPC
* Fixed coordinate frame issues and crashes
* Prepared to follow position in the 1st vehicle in map
* Fixed height map alignment
* Fixed large map and terrain tiles alignment
* Fixed inputs for new model
* Added scale factor
* Fixed slow read write operations
* Bug fixes
* Removed debug output
* Removed nvidia profinling marks
* Preparing merge debug code removed, runtime working by rounding the loading data properly
* Made pytorch optional module
* Added TRACE_CPUPROFILER_EVENT_SCOPE to Update and UpdateTexture
* Added optional pytorch conditions and macros
* Removed static Path
* Fixed #ifdef clause. Added missing include.
* Deformation on vehicle working. Ready to start optimisation
* First optimsation done. Heightmap per Tile
* Updated tiles'heightmap to make their size relative to texture and tilesize
* Fixed slow frames. Added functionality
* Removed unecessary library links
* Fixed pytorch link
* Limited search to relevant tiles
* Added multithreaded particle search. Fixed cuda architectures compilation
* Deformation plane added
* Added missing resources
* Updated particle movement update methods
* Spawn on runtime deformation plane
* Added cachemap lock when initializing a region
* Added UHeightMapDataAsset to hold heightmap data
* Fixing some settings
Co-authored-by: Aaron <samaniegoaaron112@gmail.com>
Co-authored-by: bernatx <bernatx@gmail.com>
* RssSensor fixes
RssSensor:
- make client side calculations threaded to ensure it is not
incrementally delaying the whole client
- support multiple instances of RssSensor
- initialize _last_processed_frame properly
RssRestrictor:
- enable python set_log_level
General:
- Update to RSS v4.4.2 and map v2.4.5_hotfix supporting elevation
- manual_control_rss: enalbe dynamics log level switching
* Update to ad-rss-lib 4.4.3
* really go to ad-rss-lib 4.4.3
* Changes in constant values
Added new constants
* Added Ambulance and Firetruck as unsafe vehicles in spawn_npc.py
* WIP: Rerouting algorithm
* Removing rerouting algorithm.
Fixed collisions at high speed.
* Changes to constant values and cleaning up.
* Users can now set their own path by giving a list of locations.
Implementation of RoadOption for SimpleWaypoints
Implementation of GetNextAction and GetActionBuffer for TM actions
Parameters for setting the path import and starting the import process.
Other minor changes
* Users now only need to use one parameter to import path
* added road option to the cache map
* Introduced RoadOption in InMemoryMap for each swp
Introduced getter in PythonAPI for high level path
Introduced getter in PythonAPI for next high level action
Introduced parameters for random lane changes
Fixed keep right parameter
* Introduced parameter SetImportedRoute to set a RoadOption path for TM controlled vehicles.
Method in Localization Stage to compute a path using RoadOptions.
* CHANGELOG.
* Changing names of methods
Cleanup
* Small changes to Collision Stage
Cleanup of SetRoadOption
Added debug_helper in Loc Stage (to be removed)
* Changes to get_next_action and get_all_actions
Fixed lane changes representation in those methods
Now get_next_action and get_all_actions need the Actor in Python
* Remove debug_helper
Set MAX_BRAKE to 0.7
* Constant values in Loc stage
* Code fixes
* Small changes in collision stage
Adding constants for VLStage
Changes in VLStage
* Choose randomly a right or left lane change if forcing both.
* Reviewable changes
Changes name of variables in Loc stage
Removed unnecessary import
* Minor final changes
* Change of method name in generate_traffic
* Change in Latest ContentVersion.txt
Co-authored-by: Joel Moriana <joel.moriana@gmail.com>
* added vehicle lights stage to the traffic manager under the dev branch
* update CHANGELOG to lastest version and added vehicle lights stage for automatically turn on-off lights
* added vehicle lights stage to the traffic manager under the dev branch
* removed commented code
* cleanups to the code, adding some comments to explain how the trafficmanagerlocal handles the command buffer
* fixing missing reference in ALSM constructor, disabling vehicle light state for dormant vehicles
* bugfix in vehicle_id_list access, added fix to avoid blinking brake lights due to throttle control
* requested changes in the vehicle light stage code
* missing namespace qualifier for constants
* Added explicit method to enable the vehicle lights management by the TM, on a per-vehicle basis. Changed generate_traffic.py to use the new vehicle lights stage instead of just setting a fixed, constant vehicle light status. Docs are updated, accordingly.
Co-authored-by: npunito <npunito>
Co-authored-by: Jacopo Bartiromo <32928804+jackbart94@users.noreply.github.com>
- Increase max throttle to manual control to 1.0
- In manual control and generate traffic, ignore the generation
if the filter only returns one blueprint
* Fixed typo in code snippet. Swapped order of altitude and azimuth
* Fixed image links that needed '../' prefix to be visible on the live documentation
* Removed reference to velocity being affected 2 frames after it is set
* Updated Python API with missing 'set_pedestrians_cross_factor' method
* Regenerated python_api.md
Guillermo