18 Sep
Isaac Sim 7.0.0 Alpha 1 is tagged on GitHub (tool · 1 minute read)
Isaac Sim has tagged release v7.0.0a1, its 7.0.0 Alpha 1, on GitHub.
Tag
26 items · newest first
open-source-tools
Code, models and tools released under open licences.
17 Sep
Isaac Lab 3.0 Early Access puts one task API across physics and rendering backends (tool · 8 minute read)
Isaac Lab 3.0 Early Access brings one task API across multiple physics, rendering and visualization backends, so supported tasks can switch between them without rewriting the environment. Many workflows now run kit-less, without installing or launching Isaac Sim, and a new isaaclab command replaces library-specific training and playback scripts. Quaternions change from WXYZ to XYZW, and General Availability is targeted toward the end of October 2026.
9 Sep
MuJoCo 3.13 adds a discrete integrator that keeps stiff springs stable (tool · 2 minute read)
MuJoCo 3.13.0 writes the .mjz root file as model.xml, which survives file renaming, and lets site geometries be associated with meshes. A new discrete integrator merges the constraint solve and the implicit velocity update into one operation, bringing joint, tendon and actuator stiffness into the solver's metric. Passive springs and actuator position gains stay stable at timesteps far beyond the explicit stability limit, and Python 3.15 is now supported, GIL and free-threading alike.
8 Sep
Robotics with Conan: consuming ROS as a regular package (4 minute read)
Conan's team shows how to consume ROS as an ordinary Conan package, rather than installing it system-wide with apt or rosdep on every developer machine and CI agent. It addresses the usual split where C++ control and perception code is managed with Conan while ROS sits separately on top of it. Conan's team has been building the Kilted ROS distribution from source as a regular package for Linux, macOS and Windows, so adding ROS to a C++ project becomes one more requires.
— blog.conan.ioopen-source-tools
7 Sep
Genesis 1.4 exports scenes and trajectories for bit-exact replay (tool · 3 minute read)
Genesis v1.4.0 can export a scene and a complete trajectory as a standalone archive that replays bit-exactly on any machine, aimed squarely at making bug reports reproducible. Only the rigid solver is supported so far, and the static description of entities is now separated from the runtime getters. The breaking changes cover external wrenches applied at any point on a link and morph pose offsets in the velocity and acceleration getters, alongside a long list of asset-parsing fixes.
4 Sep
NVIDIA Is Buying Hugging Face. One of Its Hottest Robots Runs on a Chinese Chip (4 minute read)
NVIDIA's proposed $12.9 billion acquisition of Hugging Face is being framed as a software deal, but Hugging Face also owns Pollen Robotics. Its new $399 Microduck, which passed 10,000 orders, runs on Rockchip's RK3566 rather than an NVIDIA processor. Pollen, the French company Hugging Face bought in 2025, lists the RK3566 and its AI accelerator among Microduck's key specifications, and the backlog has pushed some deliveries past the original Christmas target.
3 Sep
BRIDGE: An Open-Source Humanoid Platform via Morphology-Control Co-Design for Physical AI (paper · 15 minute read)
Bridge is an 88 cm open-source humanoid whose body and controller were optimised together rather than one after the other, released with its control policy. The paper introduces a metric that scores a design on both kinematic retargeting fidelity to human motion and dynamic tracking performance, and reports better numbers than Bumi, K1 and Toddlerbot on all of them.
2 Sep
AGIBOT Releases AGIBOT WORLD 2026 Theme 3 for Real-World Robot Reinforcement Learning (4 minute read)
AGIBOT has open-sourced 11,430 real-world trajectories across 14 industrial and household tasks, including 1,024 successful and 1,369 failed policy rollouts. Annotations cover task progress, completion, errors, disturbances and human interventions, giving structured execution feedback rather than demonstrations alone. It is the reinforcement-learning theme of AGIBOT WORLD 2026, an open-source dataset initiative organised into themes, and the data spans expert demonstrations, autonomous deployment and human-in-the-loop correction.
1 Sep
Show HN: I kept finding robot datasets with skewed clocks, so I built a checker (tool · 9 minute read)
embodied-sync is an open-source Python library for aligning and validating multi-sensor robot datasets, handling different sample rates, clock skew and drift, late packets and missing frames. It fits around UMI, LeRobot, ROS 2 and MCAP, and the author describes the project as alpha. The pitch is that cameras run at one rate and robot state at another, packets arrive late, and a device reconnect can silently reset its clock, so a policy trains on observations that never lined up.
31 Aug
Hebbian Robotics open-sources HFlow, an SDK for robot data pipelines (tool · 7 minute read)
Hebbian Robotics, part of Y Combinator's summer 2026 batch, has open-sourced HFlow, an SDK for multimodal data pipelines in robotics. It checks recordings for frozen cameras, drifting streams, missing topics and duplicates, runs transformations and labels, and stamps each processed episode with its provenance in a queryable catalogue, with MCAP as its input and output format.
28 Aug
ViFailback's 5,202 episodes of robot failures, now in LeRobot format (dataset · 1 minute read)
ViFailback, the dataset from a CVPR 2026 paper on learning from manipulation failures, now has a LeRobot conversion: 5,202 real ALOHA dual-arm episodes across 100 tasks, 4,545 of them failures, each failure with an annotated keyframe to avoid and one to correct. It is MIT-licensed; the depth maps were left out.
27 Aug
Hugging Face is selling a cute $399 open source duck robot, Microduck (2 minute read)
Hugging Face unveiled the Microduck, a $399 open-source robot that ships before Christmas. The 25-centimetre duck waddles, picks things up with its beak, rights itself after a fall and roller skates, following the Reachy Mini in a hardware line the company started by acquiring Pollen Robotics. It perceives with a camera, lidar and two inertial measurement units.
27 Aug
How to Train a Cross-Embodiment Robot Navigation Policy with AI Agents (11 minute read)
NVIDIA describes an agent-driven workflow in which a coding agent validates dependencies, prepares assets, runs smoke tests, launches training and compares checkpoints for a navigation policy. Human approval gates control scene acceptance, the smoke test and checkpoint promotion, so the work is not repeated for every robot-scene pair. The tutorial uses Spot as the reference robot across a built-in scene and a SAGE-10K scene, with Omniverse NuRec supporting captured environments.
26 Aug
Perceptron releases Isaac 0.5, an open-weight vision model for factory robots (3 minute read)
Perceptron, founded by two former Meta FAIR researchers, has released Isaac 0.5, an open-weight model meant to help vision-guided robots perceive, reason and act in warehouses and factories. The company says it was trained on a million hours of general video plus egocentric and UMI footage; it has raised $21 million, according to PitchBook.
21 Aug
Reactor open-sources a world model and its training code (1 minute read)
Reactor says it is releasing its world model and the training code as open source, with a blog post on the lessons it learned building them.
20 Aug
MuJoCo 3.12 moves MJCF into a single schema and adds a PID actuator (tool · 6 minute read)
MuJoCo 3.12.0 defines the MJCF grammar in a single schema file, generating the parser's grammar table, keyword maps and typed attribute bindings from it. The release also adds a PID actuator and, as a breaking change, drops the custom binary texture format in favour of PNG and KTX. The new PID actuator takes real position and velocity setpoints with optional integral action, anti-windup and slew limiting, subsuming the old plugin with proper activation state.
20 Aug
Keeping the Franka Emika Panda alive: a ROS 2 stack with a reliable position interface (paper · 15 minute read)
The Franka Emika Panda's external position interface has been unreliable for years, and this paper traces the vibrations and protective stops to the timing of the external control loop and sampling jitter rather than to the robot. The open-source stack decouples real-time communication from the ROS 2 control loop and adds rate matching and position-domain reference generation. It is validated across motion planning, compliance control, position-controlled manipulation and haptic teleoperation, on two independent Panda platforms.
18 Aug
IG-10K: 20,000 paired human-robot episodes for the Imitator Game benchmark (dataset · 1 minute read)
The IG-10K dataset accompanies The Imitator Game benchmark, with over 20,000 paired human-robot episodes across more than 50 tasks and six domains. It spans four levels of imitation difficulty in real and simulated settings, in LeRobot format.
14 Aug
Accelerating Large-scale Bundle Adjustment for LiDAR Mapping via Parallel Computing (paper · 15 minute read)
The first fully parallel framework for LiDAR bundle adjustment, built for maps larger than a GPU's memory: asynchronous data loading, bottom-up voxelisation to extract planar features, and a parallel increment solver. On large public datasets it runs up to ten times faster while matching the mapping accuracy of existing methods. The code is on GitHub.
11 Aug
TLabel – A unified annotation format and converter for tactile robotics datasets (tool · 3 minute read)
TLabel is a cross-sensor annotation schema for tactile manipulation data, producing 14-dimensional semantic annotations that different sensor types can all emit. Each adapter declares which dimensions it can and cannot annotate, so results from different tactile sensors become comparable rather than sensor-specific. Compliance levels L1 to L4 let each sensor participate at whatever information density it can actually support.
11 Aug
RynnValue: Scaling Robotic Value Foundation Models with Temporal Distance (paper · 15 minute read)
RynnValue is an open-source value model for robot manipulation trained on temporal distance to the goal, which timestamps supply for free, over more than 7,000 hours of clips. It beats the preference-supervised state of the art on RBM-EVAL-OOD, 0.675 to 0.655, and as a dense reward lifts real-world policy success from 52.5% to 72.5% online.
7 Aug
Genesis 1.3.2 makes rigid-body simulation deterministic on CPU and GPU (tool · 1 minute read)
Genesis v1.3.2 makes rigid-body simulation behave consistently across scene orientations and scales, and improves numerical robustness when the mass matrix is ill-conditioned. Simulation is now fully deterministic on a given machine, on both CPU and GPU. It also adds terrain-height querying, and fixes viewer picking, raycast sensors missing hits, and a crash on delayed sensor reads.
4 Aug
Tactus: Open-Vocabulary Object Recognition from Low-Cost Pressure Arrays (paper · 15 minute read)
Tactus answers text queries about objects from the readings of cheap resistive pressure arrays alone, reaching 0.771 top-1 on the STAG benchmark without a trained classifier head, level with the dataset's supervised CNN. The weights and code are released, and so are the failures.
3 Aug
LeRobot v0.6.1 renames a core module and loads safetensors straight to the device (tool · 7 minute read)
LeRobot v0.6.1 renames the lerobot.types module to lerobot.lerobot_types, a breaking change from v0.6.0. The release also loads safetensors directly onto the target device and improves the vision-language subtask annotation tooling. Other changes resolve policy components by convention and aggregate policy sub-losses through the metrics tracker.
3 Aug
Open-DiffLoco: Open-Source Differentiable Learning for Deployable Blind Quadruped Locomotion (paper · 15 minute read)
Open-DiffLoco is an open-source framework that trains blind quadruped locomotion policies with differentiable simulation, running the Short-Horizon Actor-Critic algorithm in MuJoCo XLA. Deployed on a Unitree Go2, the proprioceptive policy tracks omnidirectional velocity commands with under 0.2 m/s of error, runs faster than 1 m/s and withstands lateral pushes, without reference trajectories or complex auxiliary rewards. Training takes 20 to 60 minutes on a single RTX 5080 with under 6 GB of video memory.
3 Aug
SpatioLM: Towards General Physical Spatial Intelligence in Vision-Language Models (paper · 15 minute read)
SpatioLM adds a plug-in spatio-vision module to vision-language models, supervised with pseudo depth and camera information rather than 3D inputs, and reports 71.6 on VSI-Bench, the first score above 70, with transfer to manipulation. The code is on Xiaomi Research's GitHub.