AgiBot’s Expedition A2 Completes 106km Trek, Setting New Bar for Humanoid Endurance

The global humanoid robotics sector has long been defined by short, scripted laboratory demos or treadmill tests. This week, Shanghai-based AgiBot (Zhiyuan Robotics) attempted to shift that paradigm by taking its hardware onto the open road.

Between November 10 and November 13, the company’s Expedition A2 platform completed a continuous, autonomous trek of 106.286 kilometers (approximately 66 miles) from Suzhou to Shanghai. The achievement, verified by Guinness World Records as the "Longest journey walked by a humanoid robot," serves as a significant stress test for the durability and energy management systems required for the commercial labor market.

This endurance record comes as AgiBot continues to accelerate its commercialization efforts, following reports that the company is planning a Hong Kong IPO next year.

Beyond the Treadmill

Unlike many previous endurance benchmarks, AgiBot’s expedition took place in an unstructured, real-world environment.

The route required the bipedal robot to navigate a variety of surface topographies that typically plague robotic balance algorithms, including asphalt roadways, slippery tiled pavements, and tactile paving designed for the visually impaired. Furthermore, a significant portion of the trek was conducted overnight.

The "Hot-Swap" Advantage

Perhaps the most technically significant aspect of the journey was how the robot managed its power. The walk was executed without powering down, a feat made possible by a proprietary "quick hot-swap" battery architecture.

While competitors often choose integrated structural battery packs to maximize density, AgiBot has opted for a modular approach. Technicians were able to replace the energy module while the robot remained active, likely powered by an internal buffer during the interval.

This capability directly addresses the "downtime problem" in industrial automation, potentially allowing for near-100% uptime in 24-hour production cycles. A problem UBTECH also has addressed with their autonomous battery swapping and Figure with the inductive charging capabilities of the Figure 03.

Mechanical Durability

Following the conclusion of the walk at the historic Bund in Shanghai, AgiBot executives noted that the robot remained in good condition. The primary visible wear was restricted to the rubber layer of the foot soles.

This suggests that the company's proprietary "PowerFlow" actuators and transmission systems successfully withstood hundreds of thousands of load cycles without developing significant mechanical fatigue or backlash. The 175cm, 55kg robot is built with over 49 active degrees of freedom, designed to strictly mimic human anthropometry for compatibility with existing workspaces.

The "China Speed" Context

AgiBot’s rapid ascent from its founding in February 2023 to this record-breaking demonstration underscores the velocity of the Chinese robotics ecosystem. The company was founded by Peng Zhihui, a former Huawei "Genius Youth" program alumnus. As we detailed in our profile on Peng’s rise from tech influencer to entrepreneur, his celebrity status has helped the company attract top-tier talent and capital.

The company has already produced over 1,000 units as of late 2025 and is expanding production capacity in Shanghai’s Lingang Intelligent Manufacturing Park.

While other Chinese competitors like Unitree have focused on dynamic athleticism—setting records for running speeds and backflips—AgiBot appears to be positioning the Expedition A2 as a reliable "blue-collar" worker.

Commercial deployments are already underway. AgiBot has signed contracts to deploy wheeled variants of its robots at Miaoyang Fulim Precision Machining Co. and service robots at China Mobile outlets. The company is also aggressive on pricing, with the "Youth Edition" of its humanoid platform priced at approximately $27,000 (RMB 200,000).

As the industry moves from "lab queens" to operational machines, AgiBot’s 106-kilometer march serves as a strong data point that the mechanical challenges of bipedal locomotion are being solved—shifting the bottleneck to the AI "brains" required to make these machines useful.