All updates
- Shanghai University of Engineering Science and Kepler's Bumblebee Robot Win Third Place in Express Sorting at IEIC International Embodied Intelligence Skills Competition
A team from Shanghai University of Engineering Science, using the Kepler Bumblebee robot, won third place in the express sorting category at the 2026 Global Developer Pioneer Conference and IEIC International Embodied Intelligence Skills Competition, with preparation guided by Kepler's education team. The robot is a full-size humanoid with 52 degrees of freedom, a 30 kg dual-arm payload, 8 hours of continuous operation on a 1-hour charge, and 100 TOPS of onboard computing power. The robot coffee house scene previously built jointly by the university and Kepler has completed acceptance and also serves as a platform for undergraduate lab courses and scene operations training.
开普勒人形机器人 · 微信公众号 ·
- Leju Robotics Named a National-Level Specialized, Refined, Distinctive and Innovative "Key Little Giant" Enterprise
Leju Robotics has been named a national-level specialized, refined, distinctive and innovative "Key Little Giant" enterprise. The designation is organized and selected by the Ministry of Industry and Information Technology and is the highest tier in the tiered cultivation system for such enterprises.
乐聚机器人 · 微信公众号 ·
- 29 provinces and 500 organizations join forces! National Embodied Intelligence Industry Application Industry-Education Integration Community established
The National Embodied Intelligence Industry Application Industry-Education Integration Community was established on October 31 in Harbin. It is led by Leju Robotics, Northeastern University, and Harbin Vocational and Technical University, with participation from institutions in 29 provinces and more than 500 organizations across the upstream and downstream of the industrial chain. The community adopts an organizational mechanism of enterprise leadership and council governance, focusing on talent cultivation, supply-demand matching, technological breakthroughs, and teaching resource development. Leju Robotics CEO Chang Lin said the company will open up three categories of core resources: the "Kuafu" robot body, the OpenLET open-source dataset community, and industrial production line training experience.
乐聚机器人 · 微信公众号 ·
- STIER Wins Bid for Intelligent Flexible Grinding Robot Project for Aviation Material Repair
STIER's AI flexible grinding robot has won the bid for an intelligent flexible grinding project for aviation material repair, entering the complex grinding scenarios of aviation maintenance (MRO). Addressing the issues of diverse aviation material categories, large differences in wear deformation, and difficulty in forming standardized machining benchmarks, STIER builds a full-stack solution with multimodal environmental perception, multi-source heterogeneous data analysis, intelligent decision-making, and precise execution, enabling self-analysis of workpiece features, self-generation of grinding paths, and intelligent matching of process parameters. Its PolishX flexible force-controlled actuator achieves force control accuracy of ±0.8N, adjustment response <10ms, and contact force accuracy of ±0.5% FS.
斯帝尔科技 · 微信公众号 ·
- Junrou Technology completes angel round series financing, totaling 100 million yuan
Brain-computer interface company Junrou Technology has completed a series of angel round financings: the angel round was led by Jiuhe Venture Capital, and the angel+ round was invested in by Shunwei Capital, Hillhouse Venture Capital, Agibot, and others, with the three rounds completed within one year of the company's founding totaling 100 million yuan. The company's core technology is a neural worm dynamic electrode originating from Liu Zhiyuan's team at the Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences; a planar electrode array is rolled into a fiber electrode about 196 micrometers in diameter, which crawls toward multiple brain regions under magnetic field guidance through a small hole of about 1 centimeter in the skull. The first-generation brain-computer interface system has completed some animal experiments, and an IIT is expected to launch in 2027.
九合创投 · 微信公众号 ·
- Beginner CoppeliaSim 4.9/4.10 Robot Simulation Development Course Launched
Tianyi'an has launched a beginner CoppeliaSim robot simulation development course based on CoppeliaSim 4.9/4.10, teaching the software's basic operations through five projects: a six-link robot, a four-legged linkage robot, a line-tracking and collision smart car, a Mecanum-wheel car, and an E-puck smart car. The course is originally priced at 79.8 yuan, with a 2-person group rate of 59.8 yuan and a 3-person group rate of 39.8 yuan; after purchase, it can be viewed free for 2 years. Future updates are planned to cover Simulink/Python co-simulation, robotic arm control, and trajectory planning.
古月居 · 微信公众号 ·
- Ati Motors Sherpa tug in the rain
Through a roller shutter door. Down a narrow aisle. Across a rain-soaked yard. Different Sherpa tugs, the same task: hook up a cart and keep going. Rain is just part of the route. #AMR #Robotics
Ati Motors · X ·
- IDTechEx report: Which battery technology will power the next generation of humanoid robots?
A new IDTechEx report, "Batteries for Humanoid Robots 2027-2037," predicts that demand for humanoid robot batteries will reach 10.5 GWh by 2037. Currently, NMC-based lithium-ion batteries are the most commonly used in humanoid robots, as they balance high energy density, fast charging, and cycle life; the report also evaluates solid-state batteries (about 400-500+ Wh/kg, with mass production expected in the late 2020s) and lithium-sulfur batteries (commercialization expected in the early 2030s) as future options.
Robotics & Automation News ·
- Agility Robotics says more than 30 companies are considering the Digit humanoid robot
Agility Robotics has signed a new commercial agreement with logistics company GXO to introduce the Digit at two US facilities, in Statesville, North Carolina, and Lebanon, Tennessee. Operations are expected to begin in early 2027 with the Digit 4, followed by a transition to the Digit 5.
Robotics & Automation News ·
- BMW and Embotech use Automated Driving In-Plant to move 1 million vehicles
The BMW Group and Embotech say they have used Automated Driving In-Plant technology to move a cumulative 1 million vehicles at four production sites in the UK and Germany, with roughly 3,500 vehicle movements per day and availability above 99.5%.
Robotics & Automation News ·
- Agibot Proposes Worken as a Measure of Effective Intelligent Work Hours: From Horsepower and Kilowatt-Hours to Tokens, Humanoid Robots Gain a New Ruler
At its A.I.DAY event on October 29, Agibot proposed Worken (effective intelligent work hours) as a unit of measurement. The formula is the number of active robots × autonomous working duration × output efficiency ratio. It counts only robots that are online, able to execute tasks, and meet standards; it excludes teleoperation and charging/battery-swap periods from continuous autonomous working time; and it uses the qualified output of a standard worker per unit time as one equivalent.
机器人daily · 微信公众号 ·
- VLA tactile fusion in practice: integrating force control, slip detection, and visual policies into action models
A practical approach to VLA tactile fusion that integrates force control, slip detection, and visual policies into action models. The content focuses on how these three types of signals work together with visual policies within action models; it is a methodological and practical overview, and the original text does not provide specific robot models, experimental metrics, or open-source links.
Guyuehome ·
- VLA Long-Horizon Decision-Making and Safety: Task Decomposition, World Models, and Human-in-the-Loop
VLA long-horizon decision-making and safety involves three approaches: task decomposition, world models, and human-in-the-loop. Task decomposition breaks long-horizon tasks into executable subtasks, world models are used to predict the consequences of actions, and human-in-the-loop introduces human intervention during execution to improve safety.
Guyuehome ·
- VLA Post-Training and Sim2Real: Full DAgger, Residual Reinforcement Learning, and Safe Cross-Embodiment Deployment
The content focuses on VLA post-training and Sim2Real, covering three technical approaches: full DAgger, residual reinforcement learning, and safe cross-embodiment deployment. The original text does not provide specific robot models, experimental figures, or open-source links; it discusses only at the methodological level how to transfer VLA policies to real robots and ensure safety in cross-embodiment deployment.
Guyuehome ·
- Diffusion Policy: Learning Robot Manipulation Policies via Action Diffusion
Using the Push-T tabletop pushing task as a vehicle, this piece fully reconstructs the mathematical formulation of action diffusion policies, covering cosine-schedule noise addition, the noise-prediction training objective, FiLM conditioning injection, and sliding-window execution. The author also uses public pretrained weights to complete a real inference evaluation across ten seeds.
Guyuehome ·
- VLA Fine-Tuning and Parameter-Efficient Training: A Reproducible Approach from Full SFT to LoRA
For fine-tuning VLA models, the article presents a reproducible parameter-efficient training approach spanning from full SFT to LoRA. The content covers these two training paths for developers to reproduce according to the approach.
Guyuehome ·
- Spatial Perception in VLA: Multi-View RGB-D, Camera Calibration, and 3D Action Execution
Centered on the spatial perception of VLA (vision-language-action models), this discusses three stages: multi-view RGB-D input, camera calibration, and 3D action execution. The content focuses on how VLA can complete action execution in real three-dimensional space, involving multi-camera depth perception and calibration alignment.
Guyuehome ·
- VLA Evaluation Systems: From Task Success Rate to Long-Horizon Stability and Failure Attribution
VLA evaluation systems need to expand beyond a single task success rate to two additional dimensions: long-horizon stability and failure attribution. The original text discusses this evaluation framework but does not provide specific models, versions, or experimental data.
Guyuehome ·
- SoftCon: Octopus swimming control based on bionic muscle actuation and soft-body dynamics
The SoftCon framework builds an octopus swimming control pipeline from tetrahedral soft-body dynamics, muscle constraints that contract along fibers, and traveling activation waves generated by keyframe matrices and propagation filtering, using PPO to learn coordination rather than reconstruction. The framework originates from SIGGRAPH Asia 2019.
Guyuehome ·
- PlaCo Bipedal Gait Analysis: ZMP-Constrained Trajectory Optimization and Weighted Inverse Kinematics
Gu Yue Ju analyzes the PlaCo humanoid gait pipeline: a linear inverted pendulum trajectory optimization using jerk as the decision variable keeps the ZMP inside the support polygon, cubic splines connect the swing foot, and weighted QP inverse kinematics maps the center of mass and both foot targets into 26-dimensional joint velocities, completing continuous turning walking.
Guyuehome ·
- InsertAnything Evaluated in Isaac Lab for Precision Peg-in-Hole Insertion Policies with Fingertip Force Feedback
Official weights of InsertAnything's force-conditioned insertion policy were evaluated in Isaac Lab: at 0.1 mm clearance, baseline success rates for two hole types were 96–100%. With a 5 mm initial hand offset, success rates were unaffected, but with a 3 mm hole-position observation deviation, the hexagonal hole success rate dropped to 60%.
Guyuehome ·
- G1 Dribble: End-to-End Reinforcement Learning for Humanoid Robot Soccer Dribbling Control
G1 Dribble uses end-to-end reinforcement learning to achieve humanoid robot soccer dribbling; the walking task's weights are not adjusted, and velocity commands drive both the body and the ball. The author tested it locally in MuJoCo-Warp and recorded eighteen cloud sessions, reporting touch rhythm, the distance band for re-touching the ball, and failure modes, including that a centered start steps over the ball without touching it, a first touch occurs only when offset to the side, and at 1.5 meters per second it knocks the ball away directly.
Guyuehome ·
- HACMan++: Spatially Anchored Motion Primitives for Long-Horizon Robotic Manipulation
HACMan++ targets long-horizon manipulation for six-dimensional object rearrangement, decomposing actions into three parts: selecting a primitive, anchoring a placement point on the point cloud, and continuous parameters, with unified decision-making via a per-point, per-primitive evaluation map. On this basis, the framework achieves closed-loop chaining of pushing, grasping, moving, and placing.
Guyuehome ·
- C-1N Hexapod Robot: Reproducing Insect Tripod Gait and Testing Pretrained Policies in MuJoCo
This article reproduces the C-1N hexapod robot in MuJoCo, testing the inference performance of an open-loop tripod gait and four official pretrained policies. The author quantifies the triple coordination of hexapod walking across four dimensions: phase, support margin, contact load, and residual action magnitude.
Guyuehome ·
- OpenCR-MuJoCo: Discretized MuJoCo simulation of tendon-driven continuum robots
OpenCR-MuJoCo presents a discretized simulation method for tendon-driven continuum robots in MuJoCo, modeling the continuous backbone as a rigid chain with elastic joints, where bending is produced by tendon length differences. The work uses SoRoSim reference data and the official demo to validate static accuracy, release oscillation, and closed-loop tracking.
Guyuehome ·
- PPO Enables Autonomous Balance and Navigation Control for a Bicycle Robot in PyBullet
This study takes a bicycle robot with a reaction pendulum as its subject, derives the roll dynamics and the speed-dependent instability mechanism, and uses proximal policy optimization (PPO) to train coordinated control of the steering, rear wheel, and reaction pendulum. Experiments are carried out in PyBullet: natural fall, training, balancing, navigation, and disturbance recovery, concluding with obstacle-avoidance training.
Guyuehome ·
- Analysis of Sampling and Derivative Predictive Control for Real-Time Behavior Synthesis in MuJoCo MPC
Gu Yue Ju analyzes the MuJoCo real-time predictive control framework, which uses a residual cost to unify task descriptions, with four planners—predictive sampling, gradient descent, iLQG, and hybrid ILQS—replanning at every step within the same loop. A Cartpole comparison experiment reports the convergence speed and per-step computational cost of the four.
Guyuehome ·
- Robot Joint Backlash Compensation: From Dual-Encoder Principles to Play Operator Modeling and Python Simulation
The article addresses the backlash and lost motion problem in robot joint reducers by building a dual-inertia backlash transmission model described with a Play operator, and designing a direction-reversal state machine with rate-limited feedforward compensation. It quantitatively compares three control schemes through three sets of experiments and provides a Python simulation implementation path.
Guyuehome ·
- do-mpc: Robust Model Predictive Control with Multi-Stage Scenario Trees
Using bridge crane anti-sway as the object, the article derives the composite pendulum dynamics formed by the trolley, a homogeneous suspension rod, and the payload, and presents the construction of a multi-stage scenario tree, non-anticipative input sharing, and an equally weighted expected objective. The authors conducted measurement-intermittent comparative experiments between nominal and robust controllers in the MuJoCo physics engine.
Guyuehome ·
- Neural-Fly: Online Adaptive Control for Quadrotors in Strong Winds
Neural-Fly proposes a residual aerodynamic force adaptation law that combines wind-invariant representations with online least squares for online adaptive control of quadrotors in strong winds. The method trains a network on public flight data and conducts comparative figure-8 trajectory tracking experiments with a PyBullet quadrotor in a steady wind field, demonstrating rapid adaptation enabled by domain-shift separation.
Guyuehome ·
- Explaining Ruckig's jerk-limited time-optimal online trajectory generation
The article explains Ruckig's jerk-limited online trajectory generation, covering the seven-segment profile of the triple-integrator-chain time-optimal problem, the selection of synchronization instants across multiple axes, and the closed-form solution for braking when limits are exceeded. It also presents measured tracking and target-jump replanning results for a two-joint arm in MuJoCo.
Guyuehome ·
- Why My Humanoid Controller Keeps Trembling: Troubleshooting ZMP Constraints and Contact Force Distribution
Addressing the issue of persistent trembling in a humanoid robot controller, this article investigates it from two directions: ZMP constraints and contact force distribution. It analyzes the causes of the trembling around these two types of constraint conditions, serving as an organized troubleshooting approach for humanoid robot control debugging.
Guyuehome ·
- INT8 but no speedup: the time eaten by the memory wall on RK3588
On RK3588, converting a model to INT8 does not speed up inference; the bottleneck comes from the memory wall rather than compute. The article analyzes the portion of time consumed by memory access in this phenomenon, explaining that quantization gains are limited by bandwidth rather than computational precision.
Guyuehome ·
- Foot Force Control or Position Control: A Troubleshooting Record of the Root Cause of the Humanoid Robot's Loud Landing Thuds
Addressing the issue of loud thuds when a humanoid robot lands, the investigation points to the difference in choosing between foot force control and position control. The article presents the root-cause diagnosis process for this abnormal noise in the form of a troubleshooting record, for developers to compare against their own control schemes.
Guyuehome ·
- Two weeks with GR00T N1: taking apart the DiT action head, and what cross-embodiment transfer actually transfers
After two weeks of hands-on work with GR00T N1, the author dissects its DiT action head and discusses what cross-embodiment transfer actually transfers.
Guyuehome ·
- SmolVLA in Practice: Running the Full SO-100 Workflow on lerobot
This SmolVLA tutorial demonstrates the complete workflow for running SO-100 on lerobot, showing that small models can handle such tasks. It focuses on how SmolVLA works with SO-100, covering the entire chain from environment setup to execution.
Guyuehome ·
- Unsupported Operator List and Troubleshooting Log for Running VLA on RK3588
The article documents unsupported operator issues encountered when running VLA models on RK3588 and compiles them into a list and troubleshooting log. The material provides only the title and summary; specific operators, model versions, or performance data are not given.
Guyuehome ·
- Autel Robotics Releases AAGI Air-Ground Intelligence Solution and AIP Platform
Autel Robotics launched the AAGI (Autel Air-Ground Intelligence) solution brand in Shenzhen, and together with Daohe Tongtai and Saifang Technology introduced the AIP foundation along with three industry application platforms: EIS 2.0, Avant Hub 2.0, and Skyshield 2.0.
36Kr Embodied Intelligence ·
- Press conference for Humanoid Robots and Embodied Intelligence Real-World Training Challenge held in Beijing
On November 2, the press conference for the Humanoid Robots and Embodied Intelligence Real-World Training Challenge was held in Beijing. The event is co-hosted by the Ma'anshan Municipal People's Government, the Chinese Institute of Electronics, and the World Robot Cooperation Organization, and a team assembly call was issued. The challenge features 13 competition events across 5 major scenarios—"embodied intelligence + logistics, manufacturing, service, safety, and culture and sports"—moving the "exam room" into real production and living environments such as Ma'anshan manufacturing enterprises, logistics companies, hotels, communities, and supermarkets. Ma'anshan has already built an embodied robotics innovation center, forming a capacity to supply 500,000 hours of high-quality real-machine data per year, with over 1,600P of computing power online.
36Kr Embodied Intelligence ·
- "Biweekly Chain" Session 11: Greater Bay Area Robot Full-Industry-Chain Supply-Demand Matchmaking — Embodied Intelligence Data Collection: Device Adaptation, Real-World Scenarios, and Quality & Efficiency
The 11th "Biweekly Chain" Greater Bay Area Robot Full-Industry-Chain Supply-Demand Matchmaking event will be held on November 6, 2026, in Longgang, Shenzhen, under the theme "Embodied Intelligence Data Collection: Device Adaptation, Real-World Scenarios, and Quality & Efficiency." The event is aimed at companies with robot deployment needs as well as complete-machine, component, and integration companies, with a scale of 30–50 participants and free registration, focusing on face-to-face supply-demand connections across the data collection industry chain.
深圳市机器人协会 · 微信公众号 ·