On March 3, Chinese humanoid robotics company UBTech announced the launch of the world’s first multi-unit, multi-scenario, and multi-task collaborative training for humanoid robots at Zeekr’s 5G smart factory. This initiative aims to explore group operation solutions for general-purpose humanoid robots in complex industrial environments.
Similarly, on February 27, American humanoid robotics unicorn Figure AI released a video showcasing multiple Figure 02 robots collaboratively sorting packages. The robots identified barcodes via onboard cameras, picked up items, and even self-corrected missed tasks.
From manufacturing to logistics, and from China to the U.S., humanoid robots are increasingly operating in coordinated groups across industries. This trend hinges on advancements in mass production and multi-robot collaboration capabilities.
The “Mass Production Era”: What Are the Key Drivers?
2025 is predicted to be the “mass production元年” (first year) for humanoid robots. The rise of multi-robot collaboration stems from matured individual robot technologies. Modern humanoid robots now possess basic perception (e.g., barcode recognition), motion control (e.g., object manipulation), and autonomous decision-making (e.g., error correction). Group efficiency is further enhanced by three technological pillars: high-bandwidth, low-latency communication networks; distributed computing and edge AI; and multi-modal perception and behavioral planning algorithms. These innovations are reshaping production models—shortening cycles in manufacturing and boosting sorting throughput in logistics through dynamic task allocation.
Leading Players in Multi-Robot Collaboration
UBTech – Walker S Series
At Zeekr’s factory, UBTech deployed its BrainNet software architecture and Internet of Humanoids (IoH) platform to enable group intelligence. Powered by the DeepSeek-R1 multi-modal reasoning model, Walker S robots can decompose, schedule, and collaborate on complex tasks. UBTech has partnered with automakers like BYD, Geely, and FAW-Volkswagen, with its robots undergoing single-unit training globally. The next phase will focus on multi-robot data collection and model refinement to accelerate industrial scaling.
Figure AI – Figure 02
Figure’s robots leverage the proprietary Helix Vision-Language-Action (VLA) model, enabling task execution without individual training. In a demo, two Figure 02 robots sorted groceries by visually identifying items and placing them in designated spots. The company has secured a contract to deploy 100,000 units over four years for a major U.S. client, alongside its existing partnership with BMW.
Agility Robotics – Digit
Agility’s bipedal Digit robots, capable of carrying 16 kg loads, are being tested in Amazon and GXO warehouses. Using shared environmental data, multiple Digits adjust positions dynamically to avoid conflicts. Agility’s first mass-production facility is set to roll out thousands of units by 2025, targeting annual output of 10,000.
Tesla – Optimus
Though technical details remain sparse, Tesla demonstrated multiple Optimus robots collaborating in a simulated factory in October 2024. CEO Elon Musk highlighted plans to deploy them in Tesla plants by 2025, utilizing autonomous vehicle-inspired AI for task allocation.
Sanctuary AI – Phoenix
Sanctuary’s sixth-gen **Phoenix** robots, tested at Magna’s factory, use the Carbon AI system to synchronize tasks like bolt-tightening and parts sorting. The system mimics human teamwork to enhance flexibility.
1X Technologies – Eve
Norway’s 1X employs distributed AI and wireless communication to coordinate **Eve** robots in logistics and security tasks. Backed by $136 million in funding (including from OpenAI), 1X aims to expand commercial applications.
Magic Atom – MagicBot
Launched in 2024, Magic Atom (spun off from Dreame Technology) demonstrated small-scale collaboration in assembly lines, with robots handling inspection, material handling, and scanning.
Closing Notes
While multi-robot collaboration shows promise in manufacturing, logistics, and services, challenges like synchronization, energy efficiency, and environmental robustness persist. However, as costs decline and technology matures, such systems are poised to redefine automation within the next decade.
