Dexterous Hand Practice 22 builds a closed-loop process from visually recognizing a cup to grasping it stably, connecting perception output into the control stage to complete the grasp. The content is organized around the task chain of "seeing the cup—grasping it stably" and is practical, tutorial-style instruction.
Dexterous Hand Practice 21 focuses on how VLA (vision-language-action models) control a dexterous hand, outlining the process from language instruction input to the generation of constrained actions. The content is tutorial in nature, explaining this control chain, and does not provide specific robot models, experimental data, or success rate metrics.
The 25th installment of the Dexterous Hand Practice series works backward from job posting requirements to outline a learning path for dexterous hand engineers and identify the skills the role requires. The content centers on the correspondence between job requirements and skills, and is a learning-path analysis aimed at newcomers to the field; it does not cover specific robot models or experimental data.
Tactile perception in dexterous hands mainly comes from three types of sensing solutions: joint angle, fingertip force, and slip. The article separately analyzes how joint angle sensing, fingertip force sensing, and slip sensing are implemented, and explains the role and differences of each solution in dexterous hands. The content is aimed at robotics developers and outlines approaches for selecting tactile sensors.
Robbyant has open-sourced LingBot-Touch, a tactile foundation model pretrained on about 5,000 hours of real tactile data that reads magnetic, vision-based tactile, and piezoresistive sensor inputs into a unified token space. Across 26 real robot tasks and two dexterous hands, four policies using these tokens all achieved the highest average success rates; on specific sensor models unseen during pretraining, a frozen encoder reduced contact-center error from 5.89 mm to 0.16 mm, outperforming an adapter built specifically for it.
Why it matters: The tactile foundation model unifies multiple types of tactile sensors into the same token space and provides cross-sensor and cross-task transfer results.
Professor Lan Xuguang's team at Xi'an Jiaotong University has proposed AffordSim, which connects open-vocabulary 3D affordance prediction to simulation trajectory generation. Starting from a natural language task description, it locates an object's functional regions, then combines cuRobo motion planning to generate executable manipulation demonstrations.
Gamma Intelligence, TeleAI and Unitree have jointly launched LocoManiLab, a unified evaluation framework for humanoid robot mobile manipulation. It covers five types of home spaces, 13 scenario instances and 234 tasks, with the Unitree G1 and Dex3 dexterous hands as the unified test platform.
Google DeepMind, Apptronik and Mercedes-Benz conducted a proof of concept in which Gemini Robotic 2.5 drove the Apollo 2 robot to pick and organize car parts for assembly, achieving 96% accuracy. The result is currently at the proof-of-concept stage; deployment scale and production-line conditions have not been specified.
Suzhou-based general embodied intelligence company UniX AI announced on November 9 that it has completed a RMB 500 million Pre-A funding round. Existing shareholders Xienuo Investment and Yonghua Investment increased their investment, while Greenland Group, Keli Sensing, Huagong Venture Capital, Xinjing Fuying, Redbird Qihang Fund, and Yulin Industrial Investment joined as new investors.
A team at the University of California San Diego proposed EgoThumb in an arXiv preprint, a wearable robotic thumb driven by a palm-side wide-angle camera and an imitation learning model that can autonomously grasp and stabilize objects, allowing users to complete tasks such as unscrewing bottle caps and stirring tea with one hand.
Ant Group's Lingbo on November 10 open-sourced LingBot-Touch, a tactile pretrained foundation model for dexterous manipulation that can encode different sensing signals such as piezoresistive and vision-based tactile signals into unified tactile tokens for use by different dexterous hands and policy models.
Why it matters: The original article provides comparative data on the cross-sensor tactile pretrained model across 26 real-robot tasks, which can be used to assess the actual impact of tactile representations on policy success rates.
Before procuring service robots for multi-floor scenarios, four autonomous elevator capabilities need to be verified: calling the elevator on any floor, selecting a floor inside the car, detecting door opening and safely entering and exiting after the car levels, and re-localizing after exiting the elevator and seamlessly resuming the task. The REEMAN elevator control module connects to the elevator door open/close and floor signal lines through a UWB elevator gateway box, a main control box, and a button panel; the floor signals are non-polarized to simplify wiring. Single elevators use local LoRa point-to-point communication, while multiple elevators use centralized MQTT scheduling; its Fly Boat AMR once transported 200 kg through a 1.15 m elevator at an electronics factory.
Ant Group's Lingbo has open-sourced LingBot-Touch, a tactile pretraining foundation model that uniformly encodes magnetic, visuotactile, and piezoresistive tactile signals into tactile tokens, which are then passed to the action model. Team experiments showed that when π0.5 was directly fed raw tactile readings on a zipper-pulling task, its success rate dropped from 39% to 15%; after connecting LingBot-Touch, it rose to 59%.
Why it matters: The original provides a controlled experiment showing that feeding raw tactile signals into the action model instead lowers the success rate, along with the improvement after unifying tactile tokens, which can serve as a reference for tactile representation design.
On November 10, Ant Group's Lingbo Technology open-sourced LingBot-Touch, a tactile foundation model pre-trained on nearly 5,000 hours of real tactile data. It covers three sensing principles—magnetic tactile, visuotactile, and piezoresistive—and outputs unified tactile tokens for use by different dexterous hands and action models.
Why it matters: The original text provides a unified representation across three types of tactile sensing and data from 26 real-machine tests, which can be used to assess the gains a tactile foundation model brings to existing action models.
At 2026 CeMAT ASIA, Bangqi Technology is showcasing a full-stack embodied intelligence logistics solution and a "pay-per-operation-value" RaaS model, combining standardized AMR transport with humanoid robots for end-of-line operations, coordinated by its self-developed ADS scheduling system. The solution has been validated on a production line at a leading vehicle manufacturer, improving material delivery efficiency by 30%; in a warehouse goods-to-person project, equipment achieved 7×24-hour unmanned operation, boosting capacity by 125% with 99.9% accuracy. RaaS is billed by actual workload, and the full model can be deployed in 1-3 months.
In a Shanxi project, the Rhino unmanned forklift had its radar raised by 1 m; this is not parameter tuning but a structural design for shelf compatibility. Using the unmanned forklift as an example, the article notes that robot engineering deployment failures are often due to misalignment between on-site requirements and equipment configuration, rather than inadequate algorithms.
At CeMAT ASIA 2026, Zhiwang Weilai, Yuanli Lingji and Geek+ each presented warehouse solutions that move robots from moving boxes to picking goods.
Ruiman Fly Boat logistics AMRs, with a 300 kg payload and optional 100 mm lifting, replace fixed conveyors and manual carts in receiving, put-away, replenishment, picking, and shipping. They use laser SLAM navigation, so when SKUs or capacity change, only the map software needs updating. Tasks can be triggered by WMS/MES/ERP orders, one-touch call buttons, or PLC and conveyor signals, and with elevator and automatic door modules the AMRs operate across floors and in controlled areas. In one electronics factory deployment, the AMR carried 200 kg of cargo through a 1.15 m wide elevator car without manual assistance.
Packsize and Swisslog Logistics have formed a partnership to incorporate packaging automation into overall warehouse automation design, helping customers design, sequence, and implement fulfillment systems. The two companies say the joint solution reduces touches, eliminates waste, and increases throughput, and has already been deployed in multiple global projects, integrating on-demand custom packaging, ASRS, fulfillment systems, and orchestration. Packsize emphasizes that packaging equipment is a strategic control point affecting the ROI of warehouse automation investment, not an end-of-line ancillary investment.
Berkshire Grey has released multiple enhanced capabilities driven by its unified AI software platform BG Pilot, covering Scoop unloading perception, Stride intelligent orchestration, and Core precision packing.
Nature Will has released Natura-1, a 50-billion-parameter physical general foundation model for dexterous hand control and zero-shot instruction following. The model is based on a mixture-of-experts architecture, jointly trained on language, images, video, and actions, pre-trained on more than 1 million hours of first-person human data, and runs with a 100-millisecond control loop. The company says it can autonomously execute sustained workflows and handle spatial constraints, multi-object relationships, and negation conditions.
Dyna Robotics unboxed Taku on-site at the CoRL conference and put it straight to work. Driven by Dyna-2.1, the robot completed long-horizon whole-body mobile manipulation zero-shot, with no on-site data collection or post-training. The team demonstrated it live at Booth 500 in Griffin Hall.
A paper proposes that how tactile data is represented may matter more than resolution. The authors pretrained a tactile encoder, LingBot-Touch, on various sensors ranging from 3x3 glove arrays to 240x240 optical fingertips, and tested it with 4 strategies on 26 real tasks.
Duatic has launched the DxTR general-purpose industrial robot, which uses a dual-arm plus mobile base structure and can grasp, carry, place, and drive, moving thousands of bins per day. The robot can lift 30 kg and carry 200 kg, and relies on sensing and backdrivability to collaborate safely with people. Both the software and the complete machine are developed and built in-house by Duatic in Switzerland.
VOMMI is a portable demonstration collection and learning framework that connects portable RGB demonstrations to VLA post-training through offline trajectory reconstruction and online visual-motion conditioning. R2-VO refines trajectories using sparse geometric anchors, reducing the average absolute trajectory error of body and hand streams by 24.6%; a policy post-trained only on portable demonstrations has an 18.2% lower base velocity error than a policy trained on robot-collected demonstrations, with comparable end-effector translation accuracy.
The study constructs a geometric visual pretraining dataset containing only a plane, an object, and a cube hand, with trajectories generated automatically and scenes without textures or backgrounds, so that the policy is first exposed to the geometric relationship between the hand and the object. The authors evaluate ACT on three simulated robots with five manipulation tasks each and on three real-world robot tasks; in most combinations, fine-tuning from the geometric prior yields higher success rates in early training than training from scratch, and requires only a small number of task demonstrations. The results indicate that highly simplified geometric scenes can also provide an initialization that transfers across robots and to real-world tasks.
The authors propose LAG-Fusion, a latency-aware guidance fusion framework for asynchronous multimodal diffusion policy composition that allows modality-specific policies to run at their own inference rates and provide denoising guidance whenever output is available. To keep asynchronous composition consistent, the authors derive a reference-frame rebasing rule for diffusion variables under relative action representations, so that delayed guidance is aligned before fusion. In contact-rich manipulation, the authors compose a low-frequency vision policy with a high-frequency force policy; experiments with heterogeneous modality latency show better responsiveness and task performance than synchronous fusion and a purpose-designed force-aware baseline.
NoContactNoWorries proposes a Transformer-based multimodal framework that fuses RGB-D vision with robot proprioception to infer binary contact states for hand-object interaction as a pseudo-tactile signal. The authors train a single contact prediction model on multiple objects and verify that this inferred signal can support downstream reinforcement learning agents in completing in-hand object reorientation, and that it generalizes to new objects. Both simulation and real-robot experiments validate the feasibility of inferring contact from vision and proprioception. The paper has been accepted to IROS 2026.
The authors propose DexTeleop-0, a bimanual dexterous teleoperation framework that uses fingertip tactile feedback to overlay robot-side contact corrections on human tracking commands, mitigating kinematic mapping errors and the inability of vision alone to regulate contact force. The method computes localized corrections in real time via contact-stiffness approximation and the kinematic Jacobian, and adds per-contact force tracking, soft regularization of predicted contact moments, and bounded hand-joint residuals.
The study proposes Learning-Augmented Robotic Automation, a hybrid system that integrates learned task controllers and a neural 3D safety monitor into conventional industrial workflows to automatically perform flexible cable insertion and soldering on a motor production line. Using less than 20 minutes of real-world data per task, the system ran continuously for 5 hours 10 minutes without physical fencing, producing 108 motors with a 99.4% pass rate on product-level quality inspection, a takt time close to that of human workers with smaller cycle-time variation, and solder-joint quality comparable to that of skilled operators.
Why it matters: The paper reports data from continuous operation of learning-augmented industrial automation on a real production line for 5 hours 10 minutes, with 108 motors and a 99.4% pass rate, serving as a practical reference for learned control entering manufacturing.
The authors propose an adaptive collision sensitivity framework that dynamically estimates collision force based on the effective mass (EM) of each robot link and uses it to decide whether to interrupt motion, instead of applying a fixed contact threshold to the whole robot. Experiments were conducted on a real UR10e robot with AIRSKIN sensitive skin and in simulation, supplemented by simulated experiments on a KUKA LBR iiwa using joint torque sensing. In a mock pick-and-place scenario including both transient and quasi-static collisions, production efficiency improved by more than 45% compared with the standard method.
The paper proposes agentic manipulation policies (AMP), in which a generalist AI model directly controls a robotic arm and adjusts its actions as the task progresses. The authors' reference implementation, AMP1, runs unchanged across single-arm, bimanual, and mobile manipulation.
Why it matters: The paper reports zero-shot success rates and real-world trial data for a generalist AI model directly controlling a robotic arm, which can be compared with specialist policies such as VLA and WAM.
SURE is a data-efficient vision-language-action (VLA) training framework that models spatial uncertainty with a lightweight transposed-convolutional probabilistic action decoder and uses Soft-DTW alignment supervision to accommodate temporal variation across demonstrations. On LIBERO, RoboTwin 2.0, and real manipulation tasks, SURE matches or exceeds baseline VLA manipulation success rates using only 10% of demonstration data or 20% of training steps.
ODEAction proposes a chunk-free action generation paradigm that replaces the action-chunk prediction common in VLA models with continuous action evolution. It characterizes the rate of change of the current action over evolution time using a transition field over action space; during execution, it takes the current action as the initial condition and recursively generates subsequent actions via ODE integration. The method is implemented as an action head and does not depend on a specific vision-language backbone architecture. Experiments on LIBERO, EBench, RoboTwin, and real robots show better task performance and trajectory quality than matched action-chunking baselines, with particularly notable gains on fine manipulation tasks.
The research team proposed a two-component optoelectronic soft suction cup designed for kiwifruit that uses optical signals to estimate interaction forces with real fruit within known contact intervals. On fruit not used in training, it achieved grouped cross-validation R²=0.968±0.005 (MAE=0.138 N), with a median per-fruit held-out R² of 0.973 (MAE=0.134 N).
This research proposes an inference-time obstacle-guidance framework for pretrained relative-action diffusion policies. It rolls out cumulative relative actions into a workspace trajectory, evaluates obstacle interactions along the trajectory with an SDF-based obstacle model, and during denoising propagates each obstacle response to all contributing action increments. In real-robot experiments across seven obstacle-constrained scenarios spanning four manipulation tasks, the method achieved 105/140 (75.0%) collision-free task successes, compared with 8/140 (5.7%) for absolute-action pointwise guidance. Ablations show that the temporally coupled form of relative actions yields larger gains under both obstacle models.
MDF++ is a visuo-tactile world action model that uses a diffusion backbone to learn latent representations of multisensory observations end-to-end, addressing the problem that existing world action models rely on frozen pretrained encoders whose representations are not well suited for control. The authors propose a multi-task learning strategy that leverages cross-modal correlations to learn task-relevant representations, avoiding the feature collapse caused by naive joint training. On contact-rich manipulation tasks including real-world bolt screwing and nut mating, it outperforms world action models built on frozen pretrained representations by up to 50%.
SUGARv2 proposes a unified framework for learning diverse visual humanoid loco-manipulation skills through interaction reference synthesis and a shared training pipeline. The method adapts reusable task templates to different spatial configurations to synthesize coordinated robot and object motion references at low cost, and uses reference-based imitation as a common learning objective across tasks, first training privileged teachers and then training visual policies through teacher distillation and reinforcement learning.
The study proposes a planner-verified counterfactual (CF) test: holding the grasp pose fixed, a solid obstacle is inserted into the approach corridor and its points are added to the model input, after which the planner re-derives the label, flipping it from feasible to infeasible. Across eight predictors spanning point clouds, voxels, depth, and object tokens, CF-pair AUROC falls from 0.920-0.996 in natural scenes to near chance (0.470-0.504), and this flip holds across four planners, two robotic arms, and a real-object obstacle.
The authors present RevoSim, a multimodal tactile simulator for the BrainCo Revo3 dexterous hand that jointly models 285 piezoresistive taxels and 5 visuotactile fingertips. The simulator can synchronously generate pressure, depth, RGB, and marker-displacement observations; on a single RTX 5090, its depth-conditioned marker-displacement configuration supports up to 4096 parallel environments. The authors trained in-hand manipulation and grasping policies entirely in simulation, laying a foundation for future sim-to-real transfer.