Industry
Beijing Humanoid Half-Marathon 2026: 100+ Robots Complete Full-Course Test Run
More than 100 bipedal humanoid robots completed a full overnight test run of Beijing's 21 km half-marathon course on April 11-12, one week before the April 19 race. Tiangong Ultra matched its 2025 winning time of 2 hours 40 minutes.
More than 100 bipedal humanoid robots from over 70 teams completed a full-scale overnight test of Beijing's 21.0975 km half-marathon course on April 11-12, 2026, one week before the scheduled race on April 19. The test run covered every element of the course from the start line on Kechuang 17th Street to the finish in Nanhaizi Park, simulating real race conditions including navigation, battery endurance, and emergency response protocols.
The event is the world's second humanoid robot half-marathon, following the inaugural 2025 race. This year's field is substantially larger, with entries from academic institutions including a team from Technical University of Munich, marking the first international robotics team to compete in the Beijing E-Town event. The test run results provide the clearest snapshot yet of where bipedal locomotion performance actually stands heading into 2026.
Key Stats
100+
Robots in Test Run
70+
Teams Competing
21 km
Half-Marathon Distance
2:40
Best Time (2025 Record)
The Beijing Robot Marathon: How the Event Works
The Beijing E-Town Humanoid Robot Half-Marathon is organized by the Yizhuang Economic and Technological Development Zone, a district that has become a concentration point for China's robotics industry. The course runs 21.0975 km through public streets, the same route used by the parallel human race. Roughly 12,000 human runners will compete alongside the robots on April 19, separated by physical barriers throughout the course.
Robots compete in two categories: autonomous navigation, where the robot navigates the full course using onboard sensors and AI with no human control, and remote control, where a human operator guides the robot via an interface. Autonomous entries are scored separately and are the primary technical benchmark. All competing robots must be fully bipedal with no wheeled locomotion, per the rules.
🏃 Race Format
Robots race on the same 21.0975 km public course as 12,000 human runners. Two classes: autonomous navigation and remote control. All entries must be fully bipedal, no wheels permitted.
🔋 Battery Endurance Test
Completing 21 km is primarily a battery and thermal management challenge. The 2025 winner Tiangong Ultra averaged ~10 km/h for 2 hours 40 minutes. Most competitors use battery swap or tethered power for the test phase.
The April 11-12 overnight test was a full dress rehearsal. Teams ran their robots on the actual race course at night, testing navigation systems against real urban geometry, handling curbs, crosswalks, and variable road surfaces. Emergency stop protocols were also evaluated. Organizers required teams to demonstrate safe operation in proximity to road infrastructure before the live race day.
Under the Hood: Bipedal Endurance at Race Speed
Maintaining stable bipedal locomotion for 21 km is technically harder than most factory tasks these robots are designed for. Factory work involves slow, precise manipulation in controlled environments. A half-marathon demands sustained dynamic balance at speed, on variable surfaces, with declining battery voltage affecting motor torque and control margins throughout the run.
AI-generated image
Autonomous navigation systems must map and traverse the full 21 km route using onboard LiDAR, cameras, and IMUs.
Tiangong Ultra, developed by the Beijing Humanoid Robot Innovation Center, is the defending champion with a time of 2 hours 40 minutes 42 seconds from the 2025 race. It stands 180 cm tall and weighs 55 kg, placing it near the upper end of the humanoid size class. The test run confirmed the same time, indicating stable performance at approximately 10 km/h average pace over the full distance.
The core locomotion challenge for autonomous competitors is maintaining a continuous walk-run gait that balances energy efficiency against speed. At 10 km/h, a bipedal robot is operating in a hybrid gait zone where pure walking becomes unstable but a full running gait (with both feet off the ground simultaneously) demands high control bandwidth and joint actuator performance. Most competing platforms use a fast walk with dynamic stabilization rather than a true running gait.
| Robot | Developer | Height / Weight | 2025 Finish | Navigation Type |
|---|---|---|---|---|
| Tiangong Ultra | Beijing Humanoid Robot Innovation Center | 180 cm / 55 kg | 2:40:42 (winner) | Autonomous |
| TUM Entry | Tech. Univ. Munich (China robot) | TBD | 2026 debut | Autonomous/RC |
| Alibaba entry | Alibaba / partner teams | Various | Test phase | Autonomous |
| Honor entry | Honor / partner teams | Various | Test phase | Autonomous |
| ~100 others | Various Chinese institutions | Various | April 19 | Both classes |
Who's Running: Teams and Platforms
The 2026 field is significantly broader than the 2025 inaugural race. Chinese tech giants Alibaba and Honor are participating in embodied AI trials alongside the robotics-specialist teams. The inclusion of international teams, particularly from Germany's Technical University of Munich, signals that the event is gaining recognition outside China as a legitimate benchmark for bipedal locomotion.
• Beijing Humanoid Robot Innovation Center: Operator of Tiangong Ultra, the defending champion. Government-backed research consortium focused on bipedal locomotion for urban and industrial use.
• Alibaba Robotics: The e-commerce and cloud giant has been building an embodied AI division. The half-marathon entry tests outdoor navigation, not just manipulation.
• Honor: The smartphone maker expanded into robotics hardware in 2025. Their half-marathon entry leverages mobile device vision and sensor fusion expertise.
• Technical University of Munich: The only confirmed international team, using a Chinese-made bipedal platform with TUM's own navigation and control software stack.
• 100+ additional Chinese teams: University robotics labs, startup spin-outs, and established industrial robotics firms, many competing publicly for the first time.
AI-generated image
Race operations require real-time telemetry monitoring from all competing robots for safety and timing verification.
What the Marathon Tells Us About Bipedal Progress
A half-marathon is not a task robots need to do commercially. Its value is as a benchmark, not an application. Completing 21 km of urban terrain autonomously requires solving the same problems that matter for real deployment: sustained battery life, outdoor navigation on irregular surfaces, dynamic balance recovery from minor stumbles, and thermal management of actuator systems under sustained load.
The fact that Tiangong Ultra completed the course in 2025 with a consistent time (repeated in the April 11-12 test) shows reliable, repeatable performance rather than a one-time demonstration. Consistency over two full-course runs is a more meaningful data point than any single headline time.
Key Insight
The gap between the robot winner (~2:40) and a human recreational runner (~2:00-2:30) is narrowing, but the robot field's completion rate is the more telling metric. In 2025, only a fraction of starters finished the full course. How many of the 100+ 2026 entrants complete the race will reveal how broadly capability has improved across the field, not just for the top performers.
From a commercial standpoint, outdoor autonomous navigation remains a gap for most platforms. The overwhelming majority of humanoid robot deployments in 2026 are in structured indoor factory environments where GPS-independent navigation and terrain variation are not factors. The marathon tests a genuinely different capability set.
What's Coming Next
The race itself runs on April 19, 2026 at 7:30 AM local time, starting from Kechuang 17th Street by Tongming Lake. Results will give the clearest public dataset on current bipedal endurance performance. Key metrics to watch: finish rate among the 100+ starters, autonomous navigation success rate versus remote control, and whether any competitor beats the current 2:40:42 record.
Longer term, organizers have signaled interest in expanding the event format. A full marathon distance is a possible future edition. The inclusion of international teams suggests the event could evolve into a recognized international benchmark series comparable to RoboCup, but focused specifically on bipedal outdoor endurance.
Frequently Asked Questions
How fast can bipedal robots run a half-marathon?
The current record is 2 hours 40 minutes 42 seconds, set by Tiangong Ultra at the 2025 inaugural race and matched in the April 2026 test run. That works out to approximately 10 km/h average pace. The fastest human half-marathon records are around 58-60 minutes, so the performance gap remains large. However, robot times are improving faster than human records, which have been stable for decades.
Do the robots navigate autonomously or are they remote controlled?
Both. The event has two competition classes. Autonomous robots navigate the full 21 km course using onboard sensors (LiDAR, cameras, IMU) with no human input. Remote control robots are guided by an operator. Autonomous entries are scored separately and carry more technical prestige. Most teams competing for the first time enter in the remote control class.
Why does bipedal locomotion matter for industrial robots?
Factories and warehouses are built for humans, which means stairs, ramps, narrow aisles, and uneven floors. Wheeled robots require expensive infrastructure modifications. Bipedal robots can use existing human-scale environments without retrofitting. The half-marathon tests the sustained reliability of bipedal locomotion across varied terrain, which is directly relevant to factory floor deployment.
What happens if a robot falls during the race?
Teams are responsible for their own robots. Safety crews accompany each entry. In the 2025 race, several robots fell and were manually recovered by team members, who then resumed the robot's run from the same position (at a time penalty). Autonomous fall recovery, where the robot rights itself without human intervention, is a distinct capability that only a small number of platforms currently demonstrate reliably over 21 km.
The 12-Month Outlook
The April 19 race will produce the best public dataset on bipedal endurance performance from the broadest field ever assembled. With 100+ robots, 70+ teams, and the first international competitor, the results will show whether the capability improvements seen in factory deployments have translated to sustained outdoor mobility. Tiangong Ultra's consistent 2:40 time across two full-course runs sets the benchmark every other team is measured against.
Watch for the finish rate and the autonomous navigation class specifically. A high completion rate across the broader field would indicate that bipedal endurance has matured from a specialized research result to a general capability. That transition matters for commercial deployment timelines.
The Bottom Line: Over 100 bipedal robots completed Beijing's full half-marathon test course on April 11-12, and the April 19 race will be the largest public benchmark of bipedal locomotion endurance ever conducted.
Events like this serve a purpose beyond spectacle: they generate public performance data that the industry's commercial deployments don't. Factory robot results stay inside NDA-covered partnerships. Race results are public. That transparency makes the Beijing half-marathon one of the most useful data points the industry produces.