IROS 2026 Signals the Convergence of Physical AI and Robotics

The 2026 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), held September 27 through October 1 at the David L. Lawrence Convention Center in Pittsburgh, has emerged as a pivotal moment for the robotics industry. This year’s event made one thing unmistakably clear: physical AI is no longer a research concept but a deployed reality reshaping how robots perceive, decide, and act in the physical world.

The Rise of Integrated Physical AI

Perhaps the most striking theme at IROS 2026 was the shift from treating artificial intelligence as an add-on software layer to embedding it into the very architecture of robotic systems. Astribot, a Shenzhen-rooted physical AI company, made its North American debut with the T1 humanoid robot, demonstrating what the company calls a Design for AI philosophy. Rather than building a robot body and bolting on intelligence afterward, Astribot develops its Lumo AI models, embodied operating system, and cable-driven hardware as an integrated system from the ground up.

The T1 platform, priced at 8,000 for the US market, showcased autonomous tidying of miscellaneous items into a backpack, combining real-time decision-making with dexterous manipulation of deformable objects. The company’s Lumo-2 model explores latent world-action modeling, enabling robots to reason about and execute complex physical tasks that go beyond pre-programmed motions. At a recent hackathon in Beijing, 15 teams developed more than ten functional applications on the T1 in just 36 hours, demonstrating the platform’s accessibility to developers.

Humanoid Safety Architecture Reaches New Milestones

As humanoid robots move closer to working alongside humans in industrial settings, safety architecture has become a critical differentiator. Agility Robotics and FORT Robotics used IROS 2026 to announce an expanded partnership around the safety infrastructure of Digit 5, Agility’s next-generation humanoid robot. The collaboration encompasses a three-part safety architecture consisting of a safety pendant, on-robot communications, and off-robot interfaces connecting Digit with external safety systems.

Digit 5 is being engineered as Agility’s first humanoid robot designed for cooperatively safe work at scale. Its predecessors have already logged more than 65,000 hours of operation and have been deployed at customer sites including Schaeffler, GXO, and Toyota Motor Manufacturing Canada. The joint development with FORT Robotics extends cooperative safety beyond the robot itself, establishing a broader operating framework for how humanoid robots will operate in close proximity to people in warehouses, manufacturing facilities, and logistics operations.

Robots That Can See, Think, and Adapt

Another significant development came from Inbolt, which announced 2.5 million in funding to bring real-time AI vision to industrial robots. The Paris-based startup addresses a fundamental limitation in factory automation: even in highly automated facilities, most industrial robots are effectively blind. They cannot perceive part misalignment, tooling wear, or the small variations that occur naturally on every production line.

Inbolt’s solution combines 3D vision with a hardware-agnostic AI software layer that enables robots to perceive their environment and adjust control loops in real time. The system runs natively on major robot platforms including FANUC, ABB, KUKA, Yaskawa, Comau, and Universal Robots. Already deployed on over 200 robots across 100-plus factories on three continents, Inbolt’s technology serves clients including Bosch, Beko, Flex, Ford, Stellantis, and Toyota. The funding round, led by Shift4Good with participation from Bridges Climate Transition Partners, BNP Paribas Developpement, and Ora Global, brings the company’s total funding to 4 million and will support US and APAC expansion plus entry into data center and electronics manufacturing sectors.

Tactile Intelligence: The Missing Sense

Daimon Robotics presented a compelling case at IROS 2026 that the next frontier in robotic perception lies not in vision alone but in touch. Under the theme From Touch to Intelligence, the company showcased its full-stack tactile intelligence infrastructure, including the DM-Tac series of vision-based tactile sensors with over 110,000 sensing units and more than twelve tactile modalities.

The centerpiece was the first public in-person demonstration of Daimon-TWM, the company’s tactile world model, since its August launch. The model was demonstrated through two autonomous manipulation tasks: stringing friendship bracelets with tiny engraved beads and performing heat-transfer printing on canvas totes. These tasks demand small-scale precision, deformable-material handling, and continuous adaptation to changing contact conditions, capabilities that traditional vision-only systems cannot achieve alone.

Autonomous Inspection at Industrial Scale

Beyond the conference halls, ANYbotics launched its Shift platform on October 1, representing a significant step forward in managing autonomous robot fleets for industrial inspection. Shift connects a facility’s plant model, robot fleet, inspection intelligence data, and maintenance systems into one continuous workflow. The platform is already deployed across more than 200 robot deployments and processes hundreds of thousands of inspections each month.

At a Vigier Cement plant operating around the clock, Shift has managed over 33,000 autonomous inspections across 16 months. The platform integrates with major enterprise systems including SAP, IBM Maximo, GE Vernova APM, and Yokogawa distributed control systems. ANYbotics plans to expand Shift beyond its own ANYmal robots to support third-party hardware, positioning it as a hardware-agnostic command center for heterogeneous robotic workforces.

From Individual Intelligence to Swarm Collaboration

GENISOM AI presented a vision that extends physical AI from individual robots to coordinated systems. The company showcased its Multi-Legged Collaborative Transport system at IROS 2026, demonstrating multi-robot collaboration through coordinated movement and task execution. Its embodied intelligence framework integrates locomotion, spatial, interaction, and swarm intelligence, enabling robots to perceive, understand, move, and collaborate in the physical world.

The company also debuted the Genisom L1 Maker, billed as the world’s first industry-grade educational quadruped for hands-on assembly and development. With 12 degrees of freedom, an 8 kg payload capacity, and 40-degree slope climbing capability, the platform gives students and researchers access to the complete robotics development process from assembly through algorithm development.

The Broader Industry Context

According to the International Federation of Robotics, the global market value of industrial robot installations reached an all-time high of 6.7 billion in 2026. The IFR identifies five key trends driving the industry: AI and autonomy, IT/OT convergence, humanoid reliability, new application fields, and sustainability. IROS 2026 provided concrete evidence of each trend moving from prediction to implementation.

The conference also highlighted the growing diversity of the robotics ecosystem. Faraday Future showcased its EAI Robotics World 2.0, emphasizing a one-brain multi-form multi-capability approach. Tangent Robotics raised pre-seed funding to advance fine motor skills in robots, addressing a persistent gap in manipulation capabilities. The breadth of announcements signals that physical AI is attracting investment and talent across the spectrum, from startups to established industrial players.

What Comes Next

The announcements at IROS 2026 suggest several trajectories for the coming year. First, the integration of AI models with physical hardware will deepen, with companies like Astribot and Daimon Robotics leading the way in designing systems where intelligence and embodiment are co-developed. Second, safety standards for human-robot collaboration will mature, as evidenced by the Agility-FORT partnership and new R15.08 Part 3 safety standards for industrial mobile robots. Third, fleet management platforms like ANYbotics’ Shift will make autonomous inspection a standard practice in processing industries.

Perhaps most significantly, the convergence of tactile sensing, AI vision, embodied operating systems, and swarm intelligence points toward a future where robots are not merely tools but collaborative partners capable of perceiving, reasoning, and acting in the physical world with unprecedented sophistication. The robotics industry is entering a phase where the question is no longer whether physical AI works, but how quickly it can scale.

---

Edited by Palawan @QUE.COM
Website: https://QUE.COM Intelligence
Sponsored by: https://MAJ.COM AI Autonomous