Acceleration Robotics develops hardware acceleration solutions for ROS/ROS 2-based robots, specializing in Robotic Processing Units (RPUs) via the ROBOTCORE® framework. This maps computational graphs to heterogeneous architectures (multi-core CPUs, GPUs, FPGAs) for accelerated perception, transforms, real-time DDS-RTPS/UDP/IP networking, and deterministic low-latency execution in industrial manipulation, autonomous mobility, and collaborative robotics applications.
RSF defines a common language for robot service capability, lifecycle operations, certification pathways, and service-provider networks.
Xiaomi has made significant advancements in deploying its humanoid robot on automotive production lines, achieving a 98% success rate at a self-tapping nut loading station. This improvement narrows the gap to human workers' qualification rates to just one percentage point, showcasing the robot's enhanced capabilities after four months of development. The importance of this development lies in Xiaomi's ability to expand the robot's role in manufacturing, as it has successfully taken on additional tasks such as center console side panel sorting and parts bin folding and recycling, both achieving a 90% success rate. This marks a significant milestone in the robot's operational capabilities, particularly in handling flexible workpieces, which are typically more challenging than rigid components. Looking ahead, Xiaomi's continued focus on refining its humanoid robot's perception and manipulation skills will be crucial for further integration into automotive assembly operations. No further timeline was disclosed at the time of publication.
AI and RoboticsAs the robotics industry enters a phase of large-scale development, a critical question arises: how long does it take for newly collected real-world data to translate into actionable capabilities for robots? The data journey, from collection to deployment, is complex and any delays can hinder progress. Kinetix AI is addressing this challenge by connecting every stage of data production rather than simply expanding data volume. The Kai Ego Dataset has amassed over 100,000 hours of first-person multimodal data, covering more than 2,000 atomic skills across various real-world scenarios such as homes, retail, hotels, and factories. This dataset captures the nuances of continuous tasks, allowing robots to learn complex behaviors rather than isolated actions. It integrates diverse information, including visual data, body posture, and motion semantics, providing a unified data foundation for cross-domain transfer. KAI Halo, a standardized data collection tool developed by Kinetix AI, addresses common issues encountered in real data production, such as occlusion and data quality fluctuations. By employing a four-way fisheye global shutter RGB camera and a 200Hz IMU, KAI Halo synchronizes multiple perspectives, enabling a comprehensive reconstruction of human actions and interactions with the environment. No further timeline was disclosed at the time of publication.
Embodied Intelligence Data Infrastructure Robotics AI Data ProcessingOn July 15, BeingBeyond officially launched the Being-M0.7, the world's first full-body mobile operation implicit world action model (Latent WAM). This model enables robots not only to 'see' the world but also to 'understand' how it operates, facilitating full-body movements and dexterous actions. It connects visual perception with physical interaction, allowing robots to exhibit human-like coordination. The significance of Being-M0.7 lies in its ability to learn from over 10,000 hours of human first-person video pre-training, enabling it to predict physical changes based on implicit visual and motion information. Before executing commands, the model learns three core capabilities: visual context understanding, future state prediction, and implicit representation of humanoid kinematics. This foundational understanding precedes its ability to control movements. Looking ahead, BeingBeyond's advancements in implicit world action models are noteworthy. The Being-M0.7 expands the capabilities of previous models, transitioning from desktop dexterous tasks to full-body mobile operations. Future demonstrations will likely showcase the robot's understanding of physical interactions and its ability to perform complex tasks autonomously, marking a significant step in humanoid robotics development.
Humanoid Robots Robotic Mobility AI Machine LearningResearchers from the University of Science and Technology of China have developed the OriCube, a compact six-dimensional force/moment sensor that mimics human fingertip sensitivity. Measuring just 14×14×12 mm and weighing 4 grams, it achieves a remarkable resolution of 3 millinewtons within a 23-newton range, allowing it to detect even the lightest touch, such as a feather. This innovation is significant as it addresses the limitations of current robotic tactile solutions, which often rely on electronic skin or array sensors that face challenges like complex wiring and data processing. By embedding the OriCube directly into the fingertips of robotic hands, the sensor captures minute force changes and calculates precise contact points and force vectors, offering a new approach to tactile perception in robotics. The OriCube has demonstrated low power consumption of 45 milliwatts, minimal crosstalk, and high measurement accuracy. Its ability to sense both delicate touches and withstand impacts positions it as a robust solution for enhancing robotic dexterity in uncertain environments. No further timeline was disclosed at the time of publication.
Robotics Tactile Sensors Force Sensing Artificial IntelligenceA team of PhD students born after 2000 has developed a biomimetic flapping robot capable of fluid navigation, announced by Eagle Eye Intelligent Wings. The company recently completed a Series A funding round, raising tens of millions of yuan, led by Yuanhe Puhua with participation from Futen Capital and Houxue Capital. This marks the third funding round for the company within three months since its establishment 15 months ago. The funding will primarily support the mass production of their first consumer product, the 'Eagle X,' and the development of the next-generation flapping robot and fluid simulation engine. Founded in March 2025 in Shenzhen, Eagle Eye Intelligent Wings is among the early companies focusing on embodied intelligent flapping robots. The core team consists of over ten members from Shanghai Jiao Tong University, all born after 2000, with notable achievements in research. The 'Eagle X' has completed over 3,000 hours of flight testing and is set to launch on Kickstarter in Q3 of this year. The next-generation product will feature approximately 15 degrees of freedom, allowing for independent wing adjustments. The Vortrix fluid simulation engine is expected to be opened for external use, enhancing training for flying robots and optimizing aerodynamics for fixed-wing aircraft and wind turbine blades. No further timeline was disclosed at the time of publication.
Biomimetic Robots AI Fluid Dynamics Robotics Drone TechnologyProfessor Yang Shuo from Harbin Institute of Technology (Shenzhen) has founded PHANES AI, focusing on human data, tactile perception, and world model research. The startup aims to enable humanoid robots to perform agile full-body movements. A key challenge identified is the gap in current training data, where visual cues do not capture the tactile feedback necessary for successful robotic operations. This initiative is significant as it addresses the limitations of existing robotic training methodologies, which rely heavily on visual data without incorporating tactile information. Recent studies, including NVIDIA's EgoScale, highlight the importance of first-person human operation data in training robots for complex tasks. By leveraging large amounts of human data combined with minimal real-world data, PHANES AI seeks to enhance the success rate of robots in intricate operations. Looking ahead, PHANES AI plans to develop innovative methods for collecting and scaling tactile data through its EgoTouch system, which integrates visual and tactile information. The startup's approach aims to bridge the gap between visual perception and physical interaction, ultimately improving the capabilities of humanoid robots in real-world applications. No further timeline was disclosed at the time of publication.
Humanoid Robots Tactile Perception Robotics AI Data-Centric AIOm AI has introduced the VLX series, the world's first edge-based streaming multimodal model for physical AI, designed to integrate continuous perception, precise localization, and decision-making. This innovative architecture, which emerged from years of focused development, positions Om AI at the forefront of the physical AI sector, which has seen over $6.4 billion in funding in the first quarter of 2026 alone. The significance of the VLX series lies in its ability to provide a cohesive framework that allows various physical devices, such as robots, drones, and security cameras, to transition from passive execution to active scene adaptation. This shift is crucial as the industry moves from digital AI to physical AI, with a growing emphasis on real-world applications and the need for robust, responsive systems. Looking ahead, the industry is expected to witness further advancements in physical AI, particularly as the demand for edge-based intelligent systems increases. Om AI's commitment to developing the VLX series reflects a strategic response to the evolving landscape, with no further timeline disclosed at the time of publication for upcoming milestones or product releases.
Mistral AI has launched Robostral Navigate, the first AI model specifically designed for robotic navigation. This marks a significant shift for the French company, which has previously focused on large language models, as it ventures into Physical AI. The goal is to enable robots to understand natural language instructions, interpret their surroundings using a standard RGB camera, and plan routes without relying on complex sensor infrastructures. The introduction of Robostral Navigate is important as it simplifies the navigation process, traditionally reliant on multiple technologies like LiDAR and depth cameras, which are costly and complex to integrate. By utilizing only RGB images and natural language commands, Mistral AI's approach could significantly reduce costs for robot manufacturers. An RGB camera is much cheaper than industrial LiDAR sensors, making this technology more accessible. Robostral Navigate operates on a model with 8 billion parameters, balancing computational power and operational efficiency. This size allows for faster execution on embedded platforms with limited resources, crucial for timely navigation decisions. Mistral AI trained the model on nearly 400,000 trajectories across over 6,000 simulated environments, showcasing its potential for real-world applications. No further timeline was disclosed at the time of publication.
À la une IA Industrie Robotique AMR benchmark R2R-CEAnt Group's robotics subsidiary, Robbyant, has adopted a unique approach by prioritizing the development of a robot brain over hardware design. This strategy leverages the AI infrastructure behind Alipay, which processes billions of payment transactions daily, to enhance robots' ability to perceive their environment and make autonomous decisions. The introduction of the LingBot-VA 2.0 marks a significant advancement as it is the world's first embodiment-native pre-trained VA base model, specifically tailored for robots. In real-world tests, robots equipped with this model achieved a remarkable success rate of 93.6% in dual-arm tasks, showcasing the potential of this innovative technology. As hardware becomes increasingly standardized, the focus on software intelligence and data advantages is emerging as a new competitive edge. Ant Group's initiative to create a robot brain from its extensive payment data aims to bridge the gap between the digital and physical worlds, positioning Robbyant as a leader in the robotics field.
Robotics AI Machine Learning AutomationA new framework has been proposed for tactile-based proximity sensing of buried objects using granular media jamming. This method aims to enhance the detection capabilities of robotic systems in challenging environments where traditional sensing methods may fail. The significance of this framework lies in its potential applications in search and rescue operations, archaeological explorations, and environmental monitoring. By improving the ability to sense buried objects, it can lead to more efficient and effective operations in various fields. Future developments will focus on refining the technology and testing its effectiveness in real-world scenarios. No further timeline was disclosed at the time of publication.
RESEARCH ARTICLERoboBusiness 2026 takes place October 19–21 at the Santa Clara Convention Center. Produced by WTWH Media (The Robot Report / RBR), this premier commercial robotics event covers business strategy, investment, deployment trends, and innovations across manufacturing, logistics, healthcare, and more.
A comprehensive guide to industrial robot programming language evolution: G-code and NC origins, FANUC KAREL/TP, ABB RAPID, KUKA KRL, Yaskawa INFORM, the ROS open-source revolution, ROS2 architecture, ROS-Industrial consortium, OPC-UA, TSN, and real-time Linux — how six decades of software history define today's factory floor.
A technology-first analysis of the three pillars driving industrial robot capability: precision reducers, servo systems, and motion controllers. Maps the full global supplier ecosystem — Japan, Europe, and China — from component specs to product generations and integration architectures.
Construction robotics attracted $4.4 billion in built environment tech funding through Q3 2025. SoftBank acquired ABB Robotics for $5.375 billion. RaaS is displacing capex models. AI-native platforms command 39x revenue multiples. This is where money meets machines.
Inspection, scanning, and digital twin robots are construction's widest adoption curve. BCA-mandated drone façade inspections in Singapore, DroneDeploy's 3M+ sites, Boston Dynamics Spot Cam 2, and Exodigo's $214M funding define the segment. Key stat: Inspection/scanning commands 30% of construction robot applications (FactMR, 2025; unverified).
lexQube AB (FLEXQ) secured a record USD 13M order and saw its stock rally over 150% from its 2025 low. Explore 12 months of order flow, AMR technology, and market cap growth in this in-depth RobotToday analysis.
Flexiv unveils the Enlight robotics platform, Orion controller, and MICO system—advancing force-centric, general-purpose robots for real-world applications.
D-Robotics closes $120M Series B1 to advance its chip, OS, and algorithm platform for humanoid robots, drones, and service robots—positioning itself as the infrastructure backbone of the embodied intelligence industry.
Neuracle advances implantable brain–computer interface technology in China, enabling neural control of devices and accelerating clinical applications in neurorehabilitation.
John Deere, DJI & 340+ companies compete for the $48–56B Western market and China’s $80–150B parallel ecosystem. Regional analysis across 5 top countries.
Leave a comment