Release Notes: Robotics AI Suite 25.15

Humanoid Toolkit 25.15

Humanoid Toolkit v25.15 provides necessary software framework, libraries, tools, BKC, tutorials and example codes to facilitate humanoid solution development on Intel® Core™ Ultra Series 2 processors (Arrow Lake-H), It provides Intel Linux LTS kernel v6.12.8 with Preempt-RT, and supports for Canonical Ubuntu OS 22.04, introduces initial support for ROS2 Humble software libraries and tools. It supports many models optimization with OpenVINO™ toolkit, and provides typical workflows and examples including ACT manipulation, ORB-SLAM3, etc.

New

• Provided Linux OS 6.12.8 BSP with Preempt-RT

• Provided Real-time optimization BKC

• Optimized IgH EtherCAT master with Linux kernel v6.12

• Added ACT manipulation pipeline with OpenVINO™ and Intel® Extension for PyTorch framework optimization

• Added ORB-SLAM3 pipeline focuses on real-time simultaneous localization and mapping

• Provided typical AI models optimization tutorials with OpenVINO™ toolkit

• Added pipelines:

  Pipeline Name	Description
  Imitation Learning - ACT	Imitation learning pipeline using Action Chunking with Transformers(ACT) algorithm to train and evaluate in simulator or real robot environment with Intel optimization
  VSLAM: ORB-SLAM3	One of popular real-time feature-based SLAM libraries able to perform Visual, Visual-Inertial and Multi-Map SLAM with monocular, stereo and RGB-D cameras, using pin-hole and fisheye lens models

Improved

The following model algorithms were optimized by OpenVINO™ toolkit:

Algorithm	Description
YOLOv8 model_tutorials	CNN-based object detection
YOLOv12 model_tutorials	CNN-based object detection
MobileNetV2 model_tutorials	CNN-based object detection
SAM model_tutorials	Transformer-based segmentation
SAM2 model_tutorials	Extend SAM to video segmentation and object tracking with cross attention to memory
FastSAM model_tutorials	Lightweight substitute to SAM
MobileSAM model_tutorials	Lightweight substitute to SAM (Same model architecture with SAM. See OpenVINO toolkit’s SAM tutorials for model export and application)
U-NET model_tutorials	CNN-based segmentation and diffusion model
DETR model_tutorials	Transformer-based object detection
DETR GroundingDino model_tutorials	Transformer-based object detection
CLIP model_tutorials	Transformer-based image classification
Action Chunking with Transformers - ACT model_act	An end-to-end imitation learning model designed for fine manipulation tasks in robotics
Feature Extraction Model: SuperPoint model_superpoint	A self-supervised framework for interest point detection and description in images, suitable for a large number of multiple-view geometry problems in computer vision
Feature Tracking Model: LightGlue model_lightglue	A model designed for efficient and accurate feature matching in computer vision tasks
Bird’s Eye View Perception: Fast-BEV model_fastbev	Obtaining a BEV perception is to gain a comprehensive understanding of the spatial layout and relationships between objects in a scene
Monocular Depth Estimation: Depth Anything V2 model_depthanythingv2	A powerful tool that leverages deep learning to infer 3D information from 2D images

Known Issues

• There is a known deadlock risk and limitation to use intel_gpu_top to read i915 perf event in Preempt-RT kernel, it will be fixed with next release.