Programming robotics in October 2026 centers on open ROS 2 software, tactile manipulation, and vision-language-action models. The latest company releases and research papers give developers ready-to-use stacks for perception, grasping, and humanoid movement. ROS, the Robot Operating System, is the open framework most developers use to program robots. This month's work affects warehouse, lab, and home-assistant teams because it lowers integration effort while raising demands for touch sensing, safety testing, and real-world data.
Table of Contents
- What changed for ROS developers?
- Which new arms fit research labs?
- How do robots handle boxes and fragile contact?
- What is new for humanoids in clutter?
- Key takeaways for your next build
What changed for ROS developers?
NVIDIA released Isaac ROS 5.0 on Sept. 22, 2026 at ROSCon Toronto, according to TechRepublic in its conference report. The release adds AI-agent development workflows and GPU-accelerated perception packages as free open source. It also supports ROS 2 Lyrical and Ubuntu 24.04 for about 1.3M ROS developers.
Robotiq released three open-source packages for adaptive grippers, according to Robotics and Automation News in its tooling coverage. The set includes a C++ SDK v1.0.0, an official ROS 2 ros2_control driver, and updated NVIDIA Isaac Sim assets. Developers can use the same code across ROS and non-ROS apps, then test grasps in simulation before touching hardware. Practical teams should first lock their ROS 2 and Ubuntu versions, then add one perception package at a time. Simulation assets reduce crash risk, but they do not replace payload, friction, and cable checks on the physical cell.
Which new arms fit research labs?
PAL Robotics introduced the TIAGo Pro Arm, a 7-DoF ROS 2 and ros_control arm with Series Elastic Actuators. The arm works standalone or with the TIAGo Pro OMNI Base for mobile manipulation. PAL Robotics showed the system at Booth 1005 at IROS 2026. Seven degrees of freedom means the arm can reach around obstacles much like a human arm.
Series Elastic Actuators place a spring between motor and joint, so contact feels softer and force control becomes more stable. That design suits close-range tasks such as assembly, inspection, and assisted handling. Labs choosing an arm should compare reach, payload, compliance, and ROS 2 driver maturity. A compliant research arm still needs torque limits, emergency stops, and workspace fencing before student or visitor use.
How do robots handle boxes and fragile contact?
Viam debuted BoxBot at IROS 2026 in Pittsburgh Sept. 27-Oct. 3, according to Viam via PR Newswire in the debut announcement. The research arm cuts box tape with image-based visual servoing. It then opens flaps with a vision-language-action model trained on 125 teleoperated demos in about five hours.
A vision-language-action model turns camera images and language instructions into robot actions. Daimon Robotics took a different path with Daimon-TWM, launched Aug. 10 and demoed publicly Sept. 29 at IROS. The tactile-grounded world model reasons about force, friction, deformation, and slippage to overcome vision-only manipulation limits. For builders, the lesson is direct:.
- use visual servoing to track edges, tape lines, and flap seams
- add force and slip sensing before raising speed
- collect focused demos for one package family, then expand
- log failures from glare, crushed corners, and loose tape
What is new for humanoids in clutter?
Researchers published TANGO, a whole-body vision-language-action model that let a Unitree G1 humanoid navigate clutter, narrow gaps, and obstacles. Coverage described it as a step toward home and healthcare assistance. The key shift is whole-body control, where legs, torso, and arms plan together instead of walking first and reaching later. A separate Sept. 22 arXiv paper trained scene-aware humanoid locomotion from immersive human demonstrations and deployed it on a Unitree G1. It achieved 70.2% collision-free traversal on the MTC-Challenge, including crawling and squeezing.
That score matters because crawling and squeezing remain a key limitation for real clutter. IEEE/RSJ IROS 2026 convened thousands of researchers and industry participants at Pittsburgh's David L. Lawrence Convention Center Sept. 27-Oct. 3, 2026. Workshops spotlighted Carnegie Mellon robotics and sessions on manipulation, learning, and physical AI. Teams should treat 70.2% as research progress, not deployment readiness, and test narrow-gap behavior with spotters and soft obstacles.
Key takeaways for your next build
Start with a maintained ROS 2 base, add simulation-tested grippers, then layer vision, language, and touch sensing. Use open drivers for faster prototyping, but budget time for calibration, dataset collection, and safety review. Validate each new skill on the exact boxes, doors, tools, and floors your robot will face.
You Might Also Like
- What Is New With Natural Language Processing Robotics in October 2026? Latest company releases and research papers and Key Takeaways
- What Is New With Tutorials and DIY Robotics in September 2026? Latest company releases and research papers and Key Takeaways
- What Is New With Sensors and Perception Robotics in September 2026? Latest company releases and research papers and Key Takeaways
