Players · AI-derived

AgiBot Sets Guinness Record with 110 Expedition A3 Dancing

On October 24, 2026, AgiBot's 110 Expedition A3 humanoids performed a synchronized dance in Jinan, setting the Guinness World Record for the most humanoid robots dancing simultaneously. The production units relied on LinkBots cluster control and UWB positioning before planned deployment in tourism and public services.

Video: AgiBot (official demo) — watch on YouTube

ZeroGantry analysis

The 110-unit demo confirms AgiBot can ship identical production robots that perform together for 5:43 without per-unit tweaks, a direct validation of line calibration that could cut commissioning labor by half versus mixed-fleet rivals. LinkBots fault tolerance and UWB centimeter positioning lower the barrier for 50-plus unit entertainment or service contracts, yet variable-field reliability data is still missing. Ship for controlled indoor pilots; watch real metro uptime metrics before scaling to outdoor tourism fleets.

Record-Breaking Scale in Multi-Robot Coordination

On October 24, 2026, at the Yellow River International Convention and Exhibition Center in Jinan, Shandong Province, 110 AgiBot Expedition A3 humanoids executed a 5-minute-43-second routine of synchronized dances, jumps, flips, and formation changes. A Guinness World Records adjudicator verified the count on site, awarding the title for the most humanoid robots dancing simultaneously. All 110 units were standard production models drawn from AgiBot's assembly line, with no special tuning or demo-only variants, according to partner and senior vice president Wang Chuang. This consistency stems from factory calibration of joint zero positions, torque values, and motion libraries across every robot.

The demonstration relied on AgiBot's self-developed LinkBots cluster-control platform and factory-standard ultra-wideband (UWB) positioning for centimeter-level accuracy. The system fused 5G and Wi-Fi dual-mode communication to maintain stable links across the full fleet. When any single unit experienced a dropout, the platform automatically compensated by adjusting local formations to preserve overall array integrity. Robots exited shipping crates autonomously, self-calibrated positions, and executed the full sequence without individual remote piloting. The event lasted long enough to expose drift risks in shared mapping, yet the fleet completed every transition without error.

Production Consistency and Deployment Economics

AgiBot emphasized that the 110 units represented deliverable inventory rather than bespoke hardware. Wang Chuang noted that line-level uniformity allowed the stage performance to double as a live delivery validation. This approach directly addresses a core barrier in humanoid scaling: variance between individual machines that forces per-unit commissioning time. For operators planning fleets of 50 or more units, reduced on-site calibration hours translate into lower labor costs and faster time-to-revenue in entertainment or service venues.

Post-event plans call for the same robots to enter cultural performances, tourism shows, public service roles, scientific research support, and educational demonstrations. Earlier trials already place Expedition A3 units in Jinan metro stations for directional assistance, crowd management, safety patrols, and cultural guidance. The record attempt therefore served as both capability proof and a logistics rehearsal for folding, crating, transport, and rapid unpack-deploy cycles. Each robot's collapsible design lowers shipping volume and handling labor compared with rigid-frame competitors.

Technical Architecture Behind the Swarm

LinkBots provides visualization tools for formation planning, autonomous navigation, integrated walk-run coordination, and independent multi-formation control. UWB delivers the shared time base and trajectory planning that enables millisecond-level synchronization. In the controlled exhibition hall, the system handled sunlight interference, mutual occlusion, and pedestrian movement without loss of lock. The 5-minute-43-second duration tested sustained coordination rather than single-shot choreography, revealing how distributed decision-making and dynamic obstacle avoidance scale beyond scripted playback.

Fault tolerance proved central. Automatic local compensation kept the overall pattern intact even if one or two units dropped offline momentarily. This feature reduces the need for constant human oversight during extended runs, a practical advantage when deploying dozens of units across large public spaces where network variability is common. The platform's emphasis on real-time position sensing and dynamic replanning distinguishes it from simpler motion-copying approaches used in smaller robot swarms.

Comparison with Global Humanoid Programs

AgiBot's 110-unit milestone arrives amid broader industry movement toward fleet demonstrations. Unitree has showcased coordinated G1 and H1 units in smaller groups, while Figure AI and Tesla have focused on single-unit dexterity and factory integration rather than simultaneous mass choreography. Boston Dynamics Atlas electric models emphasize dynamic balance in unstructured environments but have not yet published comparable 100-plus unit synchronized runs. AgiBot's production-volume claim—earlier reports noted over 15,000 units shipped by mid-2026—positions the company ahead in raw fleet size, yet service contracts and uptime data remain less transparent than those from Western integrators.

The record also builds on AgiBot's prior Guinness achievement: an A2 model completed a 106-kilometer cross-province walk in November 2025. That endurance test and the current swarm coordination test together illustrate a dual focus on individual reliability and collective orchestration. Both records used production or near-production hardware, reinforcing the company's narrative that demonstrations reflect deliverable capability.

Limitations and Open Questions for Fleet Operators

While the controlled indoor environment validated centimeter positioning and fault recovery, real-world tourism and metro deployments introduce variable lighting, uneven floors, and unpredictable human traffic. UWB performance under prolonged sunlight or heavy multipath conditions requires further field data. The 5-minute-43-second window, though sufficient for a record, falls short of continuous eight-hour service shifts typical in public venues. Operators will need published MTBF figures and mean-time-to-repair statistics before committing to large contracts.

Maintenance logistics for 110 units also remain unaddressed in public statements. Hot-swappable batteries helped the earlier A2 walk, but cluster-scale charging and spare-part provisioning strategies have not been detailed. If per-robot annual service costs exceed $8,000–$12,000—typical ranges cited for similar full-size humanoids—the economics of entertainment deployments tighten quickly. Cultural a

Sources

Who services this hardware

Topics

Related articles

Editorial methodology