Skip to main content

5.1.6 Version Release Notes

tros-humble​

Version: 2.4.6 (2026-02-02)​

Feature Changes (RDK S100 platform):

  • Adapted to system version V4.0.5 and OE version V3.7.0.

Version: 2.4.5 (2025-10-28)​

Bug Fixes (RDK X5 platform):

  • Fixed a bug in the image processing acceleration module hobot_cv where VSE-accelerated image resizing failed.
  • Fixed incorrect inference latency statistics in the on-device model inference framework dnn_node.
  • Fixed runtime loading failures caused by incorrect installation paths of ros component so files for MIPI image capture and algorithm modules.
  • Optimized image codec configurations: removed invalid parameters and added debug-specific configuration parameters.

Version: 2.4.4 (2025-10-24)​

New Features (RDK S100 platform):

  • The on-device algorithm inference framework dnn_node now supports ROI inference mode, where the number of output tensors equals output_size × roi_size.
  • Added a new algorithm example: Human Detection and Tracking (Ultralytics YOLO Pose).
  • Added a new algorithm example: Human Instance Tracking, which extracts human features using a reid model and stores, manages, and queries these features via an SQLite database. The launch file disables the ROI-based human tracking mode in the mono2d_body_detection node and restricts input ROI dimensions to less than 3.5× the actual model input size.
  • Added a new algorithm example: Hand Keypoint and Gesture Recognition (MediaPipe). This implements basic palm detection, pre/post-processing, and ai msg publishing; supports MIPI cameras, USB cameras, and local image replay; and supports both zero-copy and non-zero-copy image acquisition methods.
  • Added a new algorithm example: DeepSeek Large Language Model, supporting human-robot dialogue. The RDK S100 and RDK S100P platforms now support the DeepSeek_R1_Distill_Qwen_1.5B and DeepSeek_R1_Distill_Qwen_7B models.

Version: 2.4.3 (2025-09-15)​

New Features (RDK X5 platform):

Version: 2.4.2 (2025-08-29)​

New Features (RDK S100 platform):

  • MIPI Image Capture now supports the 230ai stereo module.
  • YOLO Object Detection now supports yolo11 and yolov12 algorithms; added scripts to launch via component mode, supporting 4K image inference from mipi cam.
  • The EdgeSAM Everything Segmentation algorithm now uses edgesam for model inference, supports nv12 format input, and adds padding to segmentation results for joint visualization in stereo depth examples.
  • Added a new algorithm: Text-to-Image Feature Retrieval for extracting and retrieving text-image features.
  • Added a new algorithm: DOSOD, including Diguas self-developed open-vocabulary detection DOSOD edge deployment package and a reparameterized quantization method for customizing detection categories in DOSOD models.
  • Added a new algorithm: Stereo OCC, integrating Diguas stereo OCC network.
  • The Vision-Language Model algorithm now supports the smolvlm2 model, with image replay and subscription modes; added capability to publish complete topics after llm model inference completes.
  • Added a new algorithm: Stereo Depth Estimation.

Version: 2.4.1 (2025-07-30)​

New Features (RDK X5 platform):

  • MIPI Image Capture now supports the imx415 module.
  • The EdgeSAM Everything Segmentation algorithm now uses edgesam for model inference, supports nv12 format input, and adds padding to segmentation results for joint visualization in stereo depth examples.
  • Added a new algorithm: Human Instance Tracking, which extracts human features using a reid model and stores, manages, and queries these features via an SQLite database.
  • Added a new algorithm: Stereo OCC, integrating Diguas stereo OCC network.
  • The Vision-Language Model now supports the smolvlm2 model, with image replay and subscription modes; added capability to publish complete topics after llm model inference completes.

Version: 2.4.0 (2025-05-12)​

New Features:

  • Added support for the RDK S100 platform.

Version: 2.3.3 (2025-04-30)​

New Features:

Version: 2.3.2 (2025-01-15)​

Feature Changes:

New Features:

  • Added ZED Camera Image Capture to acquire stereo images as input for stereo depth estimation algorithms.
  • Added the DOSOD algorithm, including Diguas self-developed open-vocabulary detection DOSOD edge deployment package.

Bug Fixes:

Version: 2.3.1 (2024-11-20)​

Feature Changes:

  • Upgraded the dependent opencv version from 3.4.5 to 4.5.4 (the latest release version used in Ubuntu 22.04).

New Features:

Bug Fixes:

Version: 2.3.0 (2024-09-19)​

New Features:

Version: 2.2.0 (2024-04-11)​

Feature Changes:

  • Based on TROS Foxy 2.1.3, adapted to Ubuntu 22.04 and ROS2 Humble.
  • Changed the TROS installation path from /opt/tros to /opt/tros/humble to align with ROS2’s directory structure and naming convention.
  • Discontinued the tros-ros-base installation package (which included core ROS2 packages like rclcpp, rclpy, ros2cli); standard ROS2 distribution packages are now used, and ROS2 Humble dependencies are automatically installed when installing TROS Humble.
  • Enabled ROS2 FastDDS zero-copy communication for modules handling image data, including data acquisition, image codec, and algorithm examples.
  • Changed the QoS Reliability policy for zero-copy communication from RMW_QOS_POLICY_RELIABILITY_RELIABLE (rclcpp::QoS()) to RMW_QOS_POLICY_RELIABILITY_BEST_EFFORT (rclcpp::SensorDataQoS()) to mitigate potential stability risks with zero-copy.
  • Refactored hobot_dnn to use the lower-level on-device inference framework libdnn, replacing easydnn.
  • Upgraded the speech algorithm SDK in hobot_audio to use the lower-level on-device inference framework libdnn, replacing easydnn.
  • Adapted hobot_trigger to ROS2 Humble’s rosbag2.

New Features:

  • Added Bloom compilation and packaging scripts in robot_dev_config for compiling and packaging TROS on ARM platforms.
  • Added a frame_ts_type configuration option to the hobot_mipi_cam node, supporting realtime (for communication latency measurement) and sensor (default, for sensor timestamp synchronization).
  • Added a new hobot_shm node for configuring the ROS2 zero-copy environment.

Bug Fixes:

  • Fixed compatibility issues introduced by compiler upgrades.
  • Fixed path dependency issues in on-device compilation of certain ROS2 packages.

tros-foxy​

Version: 2.1.3 (2024-03-11)​

Feature Changes:

  • Changed the data type for JPEG-compressed images from sensor_msgs::msg::Image to the standard sensor_msgs::msg::CompressedImage, enabling tools like Foxglove and ROS2 rqt to view JPEG images published by TROS. Affects modules: hobot_websocket, hobot_codec, hobot_image_publisher, hobot_usb_cam.
  • Unified JPEG/MJPEG configuration options for publishing/subscribing JPEG-compressed images; removed jpeg-compressed/mjpeg-compressed options. Affects modules: hobot_codec and hobot_usb_cam.
  • Introduced the environment variable TROS_DISTRO to indicate the TROS distribution. After running source /opt/tros/setup.bash or source /opt/tros/local_setup.bash, TROS_DISTRO is empty. Configuration file paths for modules like hobot_codec, hobot_audio, hobot_mipi_cam, and hobot_usb_cam changed from /opt/tros/lib to /opt/tros/${TROS_DISTRO}/lib.

Version: 2.1.2 (2024-01-19)​

New Features:

  • Refactored hobot_usb_cam to support more format configurations and transcoding.
  • Updated the voice SDK in hobot_audio to support both 2-mic and 4-mic microphone boards; added micphone_name configuration for device ID.

Bug Fixes:

  • Fixed incorrect step field settings in data messages sent by the hobot_rgbd_cam node.
  • Updated audio playback function calls in hobot_tts to resolve playback failures on newer system versions.
  • Removed the config device tree file from hobot_llm and updated the README; ION memory size can now be set via command-line tools on newer system versions.

Version: 2.1.1 (2023-11-03)​

New Features:

  • Added the hobot_chatbot node, integrating smart voice, large language models, and text-to-speech modules to enable on-device voice chat functionality.

Bug Fixes:

  • Fixed an issue in the text-to-speech hobot_tts node where certain characters caused the application to exit unexpectedly.

Version: 2.1.0 (2023-09-14)​

Feature Changes:

  • Updated tros-ros-base to the latest ROS2 Foxy source code, ensuring compatibility with the newest ROS2 Foxy packages.
  • When using ROS2 Foxy packages, simply run source /opt/tros/setup.bash; creating symbolic links via scripts is no longer required.

New Features:

  • The text-to-speech node hobot_tts now supports a new parameter to specify the audio playback device.
  • Added a large language model (LLM) node hobot_llm, enabling on-device LLM experiences.
  • The image codec node hobot_codec now includes a new in_format configuration option: jpeg-compressed. The subscribed topic data type is automatically selected based on this configuration.

Bug Fixes:

  • Fixed an incorrect step field setting in RGB format data messages sent by the MIPI camera node hobot_mipi_cam.

Version: 2.0.2 (2023-08-28)​

Feature Changes:

  • During tros.b installation, the configured ROS2 apt source (/etc/apt/sources.list.d/ros2.list) has been switched to Tsinghua University’s mirror, resolving slow download speeds and installation failures for ROS2 packages.

New Features:

  • Added permission checks when sourcing tros.b environment setup scripts (source /opt/tros/setup.bash and source /opt/tros/local_setup.bash). If the current user lacks root privileges, the system will automatically prompt and switch to a root account, preventing tros.b usage failures due to insufficient permissions.
  • The intelligent audio algorithm node hobot_audio now supports configuring an audio device ID parameter, facilitating secondary development.
  • The event-trigger node hobot_trigger now supports task assignment to the Trigger module via standard std_msgs topics, standardizing Trigger configuration methods.

Bug Fixes:

  • Fixed an issue where the accelerated image processing node hobot_cv failed during simultaneous crop & resize operations.
  • Resolved erroneous error logs output by the MIPI camera node hobot_mipi_cam upon startup.
  • Fixed invalid launch file configurations for the data visualization message conversion node hobot_visualization.

Version: 2.0-Release (2.0.1) (2023-06-10)​

Feature Changes:

  • Upgraded voice algorithms to improve ASR (Automatic Speech Recognition) performance.
  • Optimized the model_name configuration in algorithm examples: model_name is now automatically parsed from the model file, preventing model loading failures caused by incorrect manual configuration and improving ease of secondary development.
  • Removed the Nav2 package from the tros.b installer. Users should now directly install the latest ROS2 Nav2 package on RDK using apt, resolving stability issues present in older Nav2 versions.

New Features:

  • Added support for the RDK Ultra platform.
  • Introduced new nodes including hobot_trigger and hobot_visualization for triggering events, capturing, and visualizing rosbag data—helping users diagnose, reproduce, and visualize perception and planning/control issues in robotic scenarios. Users can also extend these nodes for custom data-triggering, recording, and real-time transmission features.
  • The USB camera capture node now auto-detects USB camera device IDs, lowering the barrier to using USB cameras.
  • Added a Visual Inertial Odometry (VIO) algorithm node, enabling low-cost, robust, high-precision robot localization based on visual inputs.
  • Added a text-to-speech node hobot_tts to convert text into spoken audio.
  • Added a LiDAR-based object detection algorithm node hobot_centerpoint.
  • Added a Bird’s-Eye View (BEV) perception algorithm node hobot_bev.
  • Added a stereo depth estimation algorithm node hobot_stereonet.

Bug Fixes:

  • Upgraded easydnn (v1.6.1) and dnn (v1.18.4) on RDK X3, fixing operator crashes and adding support for more operators.
  • Fixed incorrect depth data published by the RGBD image capture node.

Other Updates:

  • Optimized the human detection and tracking algorithm node to adaptively scale output coordinates based on input image resolution.
  • Fixed a compilation failure in the ORB-SLAM3 algorithm caused by an incorrect script path.

Version: 2.0-Beta (2.0.0) (2023-05-29)​

Version 2.0-Beta (2.0.0) is the first release in the 2.x series of tros.b. Users of tros.b 1.x are recommended to upgrade to the 2.x series.

Feature Changes:

  • Migrated the code hosting platform from GitLab to GitHub to facilitate broader community participation in secondary development.
  • Integrated a more efficient package management mechanism to accelerate version upgrades and simplify robot application deployment.

New Features:

  • Added support for the new core board development kit: RDK X3 Module.
  • Enhanced hobot_audio to output ASR recognition results, simplifying voice application development.

Bug Fixes:

  • Fixed a crash in the post-processing stage of the built-in MobileNet_SSD model within dnn_node under multi-threaded conditions.
  • Resolved model inference failures in dnn_node on x86 platforms when using DDR as input.
  • Fixed compilation failures of hobot_codec and hobot_image_publisher on x86 platforms.

Other Updates:

  • Updated example launch scripts to reference dependent modules’ launch files and configure parameters accordingly.
  • Updated the D-Robotics logo displayed on the WebSocket frontend.
Last updated on