IEEE Humanoids 2026 conference will focus on factory labor
English translation from Unibot. Source published 2026-07-27.
Humanoid robots are coming out of laboratories and into factories.
The IEEE-RAS International Conference on Humanoid Robots has closed submissions for its 25th edition. The event will take place December 6-9 in Santa Clara. The topic is “Humanoid Robots and the Future of Work,” reports Tech Times.
This choice is not an accident. For many years the conference focused on walking, balance, manipulation and control of the whole body. But the current agenda places working conditions, safety certification and economic benefits on a par with the technical program.
And here’s why: humanoid systems are already being discussed not as research stands, but as full-fledged factory and logistics equipment.
Key facts of the conference
Conference: 25th IEEE-RAS International Conference on Humanoid Robots
Dates: December 6-9, 2026
Location: Santa Clara Convention Center, California
Topic: Humanoid robots and the future of work
Expected number of participants: more than 1200 people from more than 70 countries
Submitted applications: more than 200, decisions on acceptance - until October 6
Commercial implementations set the tone for research
Tech Times points to several commercial humanoid robot programs that will define the IEEE Humanoids 2026 agenda. Boston Dynamics has already shipped the first commercial copies of Atlas to Hyundai's robotics manufacturing center and Google DeepMind in 2026. Figure AI has deployed 40 Figure 03 units at BMW's Spartanburg plant following an eleven-month Figure 02 pilot that helped produce more than 30,000 BMW X3 vehicles.
Digit from Agility Robotics also works on industrial shifts - at the Toyota Motor Manufacturing Canada plant and at the GXO Logistics warehouse in Georgia. Apptronik, meanwhile, opened Robot Park in Austin: nearly 90,000 square feet where Apollo 2 robots perform manufacturing tasks and collect training data as part of a research partnership with Google DeepMind.
Special conference coordinators Gary Bolles and Nell Watson are expected to shape the program around implementation management, workforce impact and certification. For operators, this part of the program will be no less important than the scientific reports themselves.
Technical barriers are still high, but tougher procurement questions now concern responsibilities, data rights, task allocation, and how much autonomy is allowed in actual production.
Locomotion with manipulation receives a public benchmark
The Manipulative Locomotion Competition is perhaps the most anticipated part of IEEE Humanoids 2026. 10 to 15 teams will go through a series of challenges covering mobility, manipulation, agility and logic.
Locomotion with manipulation is a combined control problem that lies between spectacular demonstrations and useful industrial work. A humanoid robot must move through a real environment, pick up objects, maintain balance during load changes, and adapt to contacts that rarely match the simulation.
The main barrier is whole-body control: robots need to coordinate gait and arm movements through a single policy, rather than processing walking and grasping as separate steps.
The competition will also compare academic robots and commercial platforms on the same tasks. Points will be awarded for autonomy, but there will also be points for teleoperation - an acknowledgment that partial autonomy is still commercially useful, and the field still needs clear public metrics of where autonomy fails.
VLA models are on the conference agenda
The article pays serious attention to vision-language-action (VLA) models as a driver of current humanoid development. In this paradigm, the robot uses visual input and language instructions to generate actions, rather than relying on narrow controllers written for a specific operation.
Figure AI's Helix model is described as a VLA architecture running on Figure 03. Google DeepMind's Gemini Robotics models are coupled with Boston Dynamics Atlas hardware, and Apptronik's Robot Park is supposed to feed real-world production data back into model training.
Bottom line: Humanoid manufacturers compete for both hardware and training data because useful generalization requires experience with a variety of physics tasks.
The program of presentations covers biomechanics, actuators, dexterous hands, model predictive control, reinforcement learning, humanoid faces, benchmarks and ethical implementation issues. Applications for workshops and tutorials, team applications for competitions and additional videos for accepted reports are accepted until August 3. Early registration closes on October 16, and final versions of articles are due on October 26.