Skip to content

Commit b8fd94d

Browse files
sgolebiewski-intelsgolebiewski-intel
committed
[DOCS] Release Notes for Embodied Intelligence SDK for 2026.0
Signed-off-by: Sebastian Golebiewski <sebastianx.golebiewski@intel.com>
1 parent fd5ceb2 commit b8fd94d

File tree

3 files changed

+85
-55
lines changed

3 files changed

+85
-55
lines changed

‎robotics-ai-suite/docs/embodied/index.rst‎

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -40,12 +40,12 @@ Humanoid - Imitation Learning Resources
4040
- :doc:`Packages List <packages_list>`
4141

4242
- :doc:`Sample Pipelines <sample_pipelines>`
43-
43+
4444
- :doc:`OpenVINO Model Optimization <openvino_optimization>`
4545

4646
- :doc:`Troubleshooting <troubleshooting>`
4747

48-
- :doc:`Release Notes <release_note>`
48+
- :doc:`Release Notes <release-notes>`
4949

5050

5151
.. toctree::
@@ -58,4 +58,4 @@ Humanoid - Imitation Learning Resources
5858
sample_pipelines
5959
openvino_optimization
6060
troubleshooting
61-
release_note
61+
Release Notes <release-notes>

‎robotics-ai-suite/docs/embodied/release-notes.md‎

Lines changed: 38 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,38 @@
1+
# Release Notes: Embodied Intelligence SDK
2+
3+
## Version 26.0
4+
5+
Embodied Intelligence SDK v26.0 introduces &pi;<sub>0.5</sub> -
6+
a Vision-Language-Action (VLA) model with open-world generalization. It is
7+
developed by [Physical Intelligence](https://www.pi.website/blog/pi05) and
8+
designed to serve as a "general-purpose AI brain" for diverse robotic hardware.
9+
The model represents a significant progress in integrating advanced reasoning
10+
with precise physical control capabilities.
11+
12+
**New**
13+
14+
- [OpenVINO](https://docs.openvino.ai) Integration: Enable &pi;<sub>0.5</sub>
15+
pipeline policy with OpenVINO inference optimized on Intel integrated GPU
16+
- Add a script for OpenVINO model conversion.
17+
- [Aloha](https://tonyzhaozh.github.io/aloha/) Pipeline Support: Include example
18+
implementation for both simulator and real robot environments.
19+
- Image Processing Optimization: Enhanced &pi;<sub>0.5</sub> model structure for
20+
improved image processing performance.
21+
- Real-Time Chunking Demo: Demonstrate &pi;<sub>0.5</sub> with Real-Time
22+
Chunking (RTC) for accelerated inference on Intel Platforms.
23+
24+
**Known Issues**
25+
26+
- `CL_OUT_OF_RESOURCES` throws an exception when running the &pi;<sub>0.5</sub>
27+
model inference with the i915 driver on an Intel Ultra 2 Platform.
28+
29+
> *Workaround: Rebind to XE driver to resolve the issue.*
30+
31+
<!--hide_directive
32+
:::{toctree}
33+
:hidden:
34+
35+
Release Notes 2025 <./release-notes/release-notes-2025.md>
36+
37+
:::
38+
hide_directive-->

robotics-ai-suite/docs/embodied/release_note.md renamed to robotics-ai-suite/docs/embodied/release-notes/release-notes-2025.md

Lines changed: 44 additions & 52 deletions
Original file line numberDiff line numberDiff line change
@@ -1,36 +1,34 @@
1-
# Release Notes
1+
# Release Notes: Embodied Intelligence SDK 2025
22

3-
Click each tab to learn about the new and updated features in each release of Embodied Intelligence SDK.
4-
5-
<!--hide_directive::::{tab-set}
6-
:::{tab-item}hide_directive--> Embodied Intelligence SDK v25.36
7-
<!--hide_directive:sync: jazzyhide_directive-->
3+
## Version 25.36
84

95
Embodied Intelligence SDK v25.36 enhances model optimization capabilities with OpenVINO™ toolkit and provides typical workflows and examples, including Diffusion Policy (DP), Robotic Diffusion Transformer (RDT), Improved 3D Diffusion Policy (IDP3), Visual Servoing (CNS) and LLM Robotic Demo. This release has also updated the real-time optimized best-known configuration (BKC) on improving AI and control performance, and supporting the Intel® Arc™ B-series graphics card (B570).
106

11-
**New Features:**
12-
13-
* Updated real-time optimization BKC, including BIOS and runtime optimization, balancing performance with AI and control consolidation.
14-
* Added support for Intel® Arc™ B-series (Battlemage) graphics card (B570).
15-
* Fixed deadlock issue when reading i915 perf event in Preempt-RT kernel.
16-
* New EtherCAT Master stack features supporting user-space EtherCAT Master and multiple EtherCAT masters.
17-
* Added Diffusion Policy pipeline with OpenVINO™ toolkit optimization.
18-
* Added Robotics Diffusion Transformer (RDT) pipeline with OpenVINO toolkit optimization.
19-
* Added Improved 3D Diffusion Policy (IDP3) model with OpenVINO toolkit optimization.
20-
* Added Visual Servoing (CNS) model with OpenVINO toolkit optimization.
21-
* Provided new tutorials for typical AI model optimization with OpenVINO toolkit.
22-
* ACRN hypervisor's initial enablement on Arrow Lake platform.
23-
* Added new Dockerfile to build containerized Robotics Development Toolkit (RDT) pipeline.
7+
**New**
248

25-
**Known Issues and Limitations**
9+
- Updated real-time optimization BKC, including BIOS and runtime optimization, balancing performance with AI and control consolidation.
10+
- Added support for Intel® Arc™ B-series (Battlemage) graphics card (B570).
11+
- Fixed deadlock issue when reading i915 perf event in Preempt-RT kernel.
12+
- New EtherCAT Master stack features supporting user-space EtherCAT Master and multiple EtherCAT masters.
13+
- Added Diffusion Policy pipeline with OpenVINO™ toolkit optimization.
14+
- Added Robotics Diffusion Transformer (RDT) pipeline with OpenVINO toolkit optimization.
15+
- Added Improved 3D Diffusion Policy (IDP3) model with OpenVINO toolkit optimization.
16+
- Added Visual Servoing (CNS) model with OpenVINO toolkit optimization.
17+
- Provided new tutorials for typical AI model optimization with OpenVINO toolkit.
18+
- ACRN hypervisor's initial enablement on Arrow Lake platform.
19+
- Added new Dockerfile to build containerized Robotics Development Toolkit (RDT) pipeline.
20+
- Added pipelines:
2621

27-
* ACRN hypervisor feature and performance
22+
| Pipeline Name | Description |
23+
|-----------------------------------------------------------------------------| -----------------------------------------------------------------------------------------------------------------------------------------------------------------|
24+
| Diffusion Policy ****diffusion_policy**** | An innovative method for generating robot actions by conceptualizing visuomotor policy learning as a conditional denoising diffusion process |
25+
| Robotics Diffusion Transformer (RDT) ****robotics_diffusion_transformer**** | A RDT pipeline provided for evaluating the VLA model on the simulation task |
26+
| LLM Robotics Demo ****llm_robotics_demo**** | A code generation demo for robotics, interacting with a chatbot utilizing AI technologies such as large language models (Phi-4) and computer vision (SAM, CLIP) |
2827

29-
#. iGPU performance degradation observed when using passthrough iGPU to VM on ACRN hypervisor.
3028

31-
#. Display becomes unresponsive in VMs when running concurrent AI workloads with iGPU SR-IOV enabled on ACRN hypervisor.
29+
**Improved**
3230

33-
**The following model algorithms were added and optimized by OpenVINO™ toolkit:**
31+
The following model algorithms were added and optimized by OpenVINO™ toolkit:
3432

3533
| Algorithm | Description |
3634
|----------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------|
@@ -42,35 +40,36 @@ Embodied Intelligence SDK v25.36 enhances model optimization capabilities with O
4240
| Improved 3D Diffusion Policy (iDP3) ****model_idp3**** | A diffusion policy model enhancing capabilities for 3D robotic manipulation tasks |
4341
| Robotic Diffusion Transformer (RDT-1B) ****model_rdt**** | A diffusion-based foundation model for robotic manipulation |
4442

45-
**The following pipelines were added:**
43+
**Known Issues**
4644

47-
| Pipeline Name | Description |
48-
|-----------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------|
49-
| Diffusion Policy ****diffusion_policy**** | An innovative method for generating robot actions by conceptualizing visuomotor policy learning as a conditional denoising diffusion process |
50-
| Robotics Diffusion Transformer (RDT) ****robotics_diffusion_transformer**** | A RDT pipeline provided for evaluating the VLA model on the simulation task |
51-
| LLM Robotics Demo ****llm_robotics_demo**** | A code generation demo for robotics, interacting with a chatbot utilizing AI technologies such as large language models (Phi-4) and computer vision (SAM, CLIP) |
45+
- ACRN hypervisor feature and performance
5246

47+
- iGPU performance degradation observed when using passthrough iGPU to VM on ACRN hypervisor.
48+
- Display becomes unresponsive in VMs when running concurrent AI workloads with iGPU SR-IOV enabled on ACRN hypervisor.
5349

54-
<!--hide_directive:::
55-
:::{tab-item}hide_directive--> Embodied Intelligence SDK v25.15
56-
<!--hide_directive:sync: humblehide_directive-->
50+
## Version 25.15
5751

5852
Embodied Intelligence SDK v25.15 provides necessary software framework, libraries, tools, BKC, tutorials and example codes to facilitate embodied intelligence 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.
5953

60-
**New Features:**
54+
**New**
6155

62-
* Provided Linux OS 6.12.8 BSP with Preempt-RT
63-
* Provided Real-time optimization BKC
64-
* Optimized IgH EtherCAT master with Linux kernel v6.12
65-
* Added ACT manipulation pipeline with OpenVINO™ and Intel® Extension for PyTorch framework optimization
66-
* Added ORB-SLAM3 pipeline focuses on real-time simultaneous localization and mapping
67-
* Provided typical AI models optimization tutorials with OpenVINO™ toolkit
56+
- Provided Linux OS 6.12.8 BSP with Preempt-RT
57+
- Provided Real-time optimization BKC
58+
- Optimized IgH EtherCAT master with Linux kernel v6.12
59+
- Added ACT manipulation pipeline with OpenVINO™ and Intel® Extension for PyTorch framework optimization
60+
- Added ORB-SLAM3 pipeline focuses on real-time simultaneous localization and mapping
61+
- Provided typical AI models optimization tutorials with OpenVINO™ toolkit
62+
- Added pipelines:
6863

69-
**Known Issues and Limitations**
64+
| Pipeline Name | Description |
65+
|-----------------------------------------------------------------------------| -----------------------------------------------------------------------------------------------------------------------------------------------------------------|
66+
| [Imitation Learning - ACT](../sample_pipelines/imitation_learning_act.rst) | Imitation learning pipeline using Action Chunking with Transformers(ACT) algorithm to train and evaluate in simulator or real robot environment with Intel optimization |
67+
| [VSLAM: ORB-SLAM3](../sample_pipelines/ORB_VSLAM.rst) | 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 |
7068

71-
* 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.
7269

73-
**The following model algorithms were optimized by OpenVINO™ toolkit:**
70+
**Improved**
71+
72+
The following model algorithms were optimized by OpenVINO™ toolkit:
7473

7574
| Algorithm | Description |
7675
|-----------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------|
@@ -91,13 +90,6 @@ Embodied Intelligence SDK v25.15 provides necessary software framework, librarie
9190
| 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 |
9291
| Monocular Depth Estimation: Depth Anything V2 ****model_depthanythingv2**** | A powerful tool that leverages deep learning to infer 3D information from 2D images |
9392

94-
**The following pipelines were added:**
95-
96-
| Pipeline Name | Description |
97-
|-----------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------|
98-
| Diffusion Policy ****diffusion_policy**** | An innovative method for generating robot actions by conceptualizing visuomotor policy learning as a conditional denoising diffusion process |
99-
| Robotics Diffusion Transformer (RDT) ****robotics_diffusion_transformer**** | A RDT pipeline provided for evaluating the VLA model on the simulation task |
100-
| LLM Robotics Demo ****llm_robotics_demo**** | A code generation demo for robotics, interacting with a chatbot utilizing AI technologies such as large language models (Phi-4) and computer vision (SAM, CLIP) |
93+
**Known Issues**
10194

102-
<!--hide_directive:::
103-
::::hide_directive-->
95+
- 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.

0 commit comments

Comments
 (0)