Robots have historically performed short, controlled demonstrations exclusively in laboratories and on stage. But now, a full-size humanoid has crossed city boundaries on public routes, achieving an endurance feat that demonstrates a new level of practical robotic readiness. A continuous, multi-day outdoor journey by a humanoid robot marks a significant milestone for embodied AI.
Agibot A2 walked about 66 miles from Suzhou’s Jinji Lake to Shanghai’s North Bund over nearly three days of continuous operation, navigating real traffic rules and unpredictable outdoor weather. The significance of this record extends beyond distance, emphasizing reliability, power management, and the robot’s ability to maintain balance for hours on unprepared surfaces.
This achievement transforms the headline into a genuine question about integrating robots into modern daily life. Where will these machines be deployed in 2026 and beyond, considering a robot can now walk autonomously from one city to another without human intervention?
To answer that, this article begins by examining the walk itself. We then trace the history of bipedal robots, from early careful steps to modern athletic agility. The article will examine how key technical advances enabled this long-trek milestone, including improvements in:
- Sensors
- Actuators
- Control software
- Batteries
We will close by previewing what this might mean for homes, hospitals, factories, and streets as humanoids begin to leave the lab for good.
Agibot A2 Endurance Record: Quick Facts and Key Statistics
- Distance and route: Agibot A2 completed approximately 106.3 kilometers from Suzhou’s Jinji Lake to Shanghai’s North Bund, formally certified as the longest humanoid robot endurance record.
- Time on feet: The journey took almost three days of continuous operation, with supervised safety procedures and traffic rule compliance along the route, as reported in mainstream coverage.
- Power strategy: Teams performed hot-swapped battery changes at regular intervals so the robot remained powered and autonomous between swaps.
- Why this matters: This record emphasizes endurance and robustness, differentiating it from earlier showcases focused on sprint speed. It complements earlier agility showcases, such as dynamic parkour by Atlas and running records by Cassie.
- Historical contrast: Earlier icons like Honda’s ASIMO delivered polished stage demonstrations and short runs, typically with limited battery windows. These feats sharply contrast with the engineering challenge of a multi-day outdoor trek.
Inside the Milestone: Agibot A2’s 66-Mile Urban Autonomy Test
From Suzhou to Shanghai: The Guinness World Record Walk
Beginning near Jinji Lake in Suzhou, the walk finished on the North Bund in Shanghai, traversing a route that blends urban streets and riverside paths. The robot moved at a modest average pace, which is expected for a continuous walk that prioritized safety and reliability over speed. The path was not a closed track.
The path involved managing curbs, ramps, uneven pavers, and the typical micro-obstacles that also challenge human pedestrians, resulting in a record focused on endurance in the wild rather than a short, controlled sprint.
Following public safety rules for slow-moving devices on shared paths, the team utilized standard precautions. These included human chaperones for crowd control and scheduled safety checks at predetermined swap points.
Late November 2025 CBS News report described a roughly three-day duration and adherence to traffic rules.
Meet Agibot A2: The Robot Behind the Record
Agibot A2 is a full-size humanoid service robot designed for public interaction, guidance, and demonstrations in spaces like malls, exhibitions, and transport hubs. While the exact configuration can vary by model, the core platform combines multi-joint legs and arms, a torso with onboard compute, and a sensor suite for perception and navigation.
Sensors and Perception
A2’s perception stack relies on multiple inputs to build a rich understanding of its environment. These systems work together to deliver real-time data for navigation, mapping the path ahead instantly:
- RGB-D cameras
- Panoramic fisheye views for situational awareness
- 3D LiDAR for depth and obstacle sensing
These sensors work together to map sidewalks, curbs, and path edges, updating the local environment in real time as people and objects move.
Brains and Actuators
Onboard compute handles real-time control and the higher-level planning. This planning keeps steps stable when surfaces change. High-torque joint actuators generate smooth, human-like gaits at low speeds that are energy efficient over long periods. The primary goal is not speed, but to maintain continuous walking without falling for hour after hour.
Power and Battery Strategy
Long outdoor walks stress energy systems more than indoor demos. During the record, the team used hot-swappable battery modules, a simple but effective way to keep the robot powered while maintaining autonomy between swaps. This method proves its utility compared with lab sprints, standing as a realistic tactic for near-term deployments in facilities and cities.
Why Endurance Matters More than Speed in 2025
While agility videos showcase physical possibilities, endurance is the critical feature that transitions robots into useful daily applications. The capability to walk for hours on unprepared surfaces makes a robot better suited to tasks like guard tours, public guidance, and maintenance checks.
In earlier milestones, Cassie captured attention with a 5-kilometer outdoor run and a later 100 meter world record, while Atlas showed athletic maneuvers and parkour in test facilities. Those achievements are real and important, yet they are still short in duration. This 66-mile walk confirms that slow, steady endurance is now a priority matching dynamic agility.
Endurance will likely be the deciding feature for the first practical assignments, as cities and companies evaluate where humanoids can be most effective. These roles include patrols in large campuses, repetitive inspections in logistics parks, and guided wayfinding at transport nodes, requiring the walker to keep moving even when interactions interrupt the rhythm.
Bipedal Evolution: From Early Humanoid Steps to Modern Athletic Agility
Honda’s ASIMO to Early Humanoid Dreams
In the 1980s and 1990s, Honda’s E-series and P-series research led to ASIMO, a small humanoid that became a global icon for friendly robotics. ASIMO could walk smoothly, climb stairs, and even jog. While ASIMO had limited battery life and its demonstrations were usually staged, it proved the achievability of stable bipedal walking alongside human-friendly gestures. Honda’s official ASIMO history documentation records the platform’s history and specifications.
The Age of Agility: Atlas, ARTEMIS, and Running Robots
In the 2010s, attention shifted to dynamic agility. Boston Dynamics’ Atlas became the dynamic agility benchmark for balance, jumps, and vaults in controlled courses. Atlas’s performances showcased advances in control algorithms, actuation, and mechanical design that allow rapid, whole-body coordination. Boston Dynamics shares these performances through robot dance routines.
University teams also pushed limits. UCLA’s ARTEMIS focused on speed and soccer-style locomotion, offering a research platform for fast gaits and kicks that connects lab progress to public-facing performance through soccer-style locomotion research.
Cassie, Guinness Records, and the Sportification of Robotics
Another branch of research explored steady running over distance and top speed in short bursts. Cassie, developed from work at Oregon State University and commercialized by Agility Robotics, set Guinness recognized marks in both distance and sprint categories. The running achievements are summarized in Cassie’s Guinness-recognized running records, which collate official notes and university releases. Competitive showcases followed, including organized races that test battery management, thermal limits, and controller robustness.
From Lab Floors to City Streets: What has Changed Since 2010
The transition from careful indoor demonstrations to real outdoor treks was made possible by four fundamental technological shifts. These changes, detailed in the sections below, have fundamentally altered the performance ceiling for autonomous bipedal locomotion:
- Better Sensors and Sensor Fusion
- Stronger, More Efficient Actuators
- Smarter Control and Planning
- Practical Power Strategies
Each of these developments addressed a major technical constraint, demonstrating that simple, reliable systems are the key to making useful walking robots.
Better Sensors and Sensor Fusion
Key components—including camera, LiDAR, and inertial measurement units—now deliver richer, cleaner data. Sensor fusion then lets the controller handle uneven pavements, ramps, and crowd dynamics without constant operator input.
Stronger, More Efficient Actuators
Improved transmissions and high-torque actuators reduce wasted energy at the joints, ensuring that slow, repetitive stepping is less power hungry for critical multi-hour tasks.
Smarter Control and Planning
Learned policies combined with model-based checks allow modern controllers to help a robot adapt to small slips, bumps, and slopes. The result is a gait that looks unremarkable to a passerby but is actually produced by thousands of micro-corrections each minute.
Practical Power Strategies
Hot-swappable batteries, conservative speeds, and rest checkpoints may seem mundane, but they are how early deployments will actually work. The 66-mile record shows that we do not need exotic breakthroughs; instead, we need reliable systems that can keep going to create useful walking robots.
China’s Robotics Strategy: Humanoid Deployment in 2026 and Beyond
A National Plan for Walking Machines
Humanoid robots are recognized as a strategic industry within China’s national planning, with public guidance setting milestones for a preliminary innovation system by 2025 and a broader world-class industry by 2027. In practice, that means funding for core parts such as actuators and sensors, shared testbeds, and pilot deployments in factories and public venues.
The strategy translates into funding for core components like actuators and sensors, along with shared testbeds and pilot deployments in public venues and factories. A formal Guiding Opinions on the Innovation and Development of Humanoid Robots outlines priorities for components, embodied AI, and testbeds. This comprehensive framework is summarized in detailed policy analysis.
At the company level, Chinese firms already span a wide range of form factors and price points. One example is an affordable, full-size platform described in an overview of Unitree’s R1. Forecasts predicting thousands of humanoid robots entering service are now intersecting with these pilots, driven by improving cost curves and endurance capabilities.
From Half Marathons to World Games: Public Testbeds
China has also turned humanoid trials into public events such as the Beijing humanoid half-marathon and the World Humanoid Robot Games. Half-marathon courses and multi-event games invite teams to prove reliability over longer distances and diverse tasks, from steady running to manipulation. Public competitions expose weak points in thermal management, battery strategy, and fall recovery.
Crucially, these events also help the public view robots as tools that must meet safety rules, rather than just as viral videos. The long walk by Agibot A2 fits into this shift from lab floors to city surfaces, where endurance and predictability matter as much as athletic flair.
The Business Case: Why China wants Humanoids in Factories and Streets
Humanoids are not the best tool for every job; for many tasks, wheeled robots remain the more efficient tool. A human-shaped machine offers a critical, unmatched advantage: it can enter spaces already designed for people. Stairs, door handles, and workstations do not need to be rebuilt.
Human-Shaped Advantage: Lowering Retrofit Costs
Because humanoids utilize existing infrastructure, they enable faster pilot programs and lower retrofit costs in warehouses, hospitals, transit hubs, and public buildings.
In healthcare, for instance, trials of hospital companions hint at roles in patient lifting, guided walking, and off-hour rounds. In retail and transport settings, service humanoids can handle wayfinding, simple inventory checks, and night-shift patrols where a steady walking pace is an asset.
A Global Race: Tesla Optimus, Apptronik Apollo, and Beyond
China’s push sits within a wider race. In the United States, Tesla has previewed its Optimus platform, with Tesla’s AI and Optimus progress detailing progress and factory ambitions. Texas-based Apptronik has positioned Apollo as a versatile industrial worker for logistics and light industrial tasks. Japan and Europe continue to fund humanoid research and safety standards. The global race in robotics creates a crowded field that will pressure prices, accelerate component improvements, and identify the most viable applications for humanoids in the late 2020s.
Humanoids in Public Spaces: Evolving Robotics for Everyday Life
From Robot Vacuums to Cross-Provincial Walkers: Everyday Robotics Evolves
Simple, autonomous helpers are already common in most homes. Tools like robotic floor cleaners, window wipers, and lawn care prove that patient autonomy, not flashy speed, is what wins in daily life.
The 66-mile walk extends this principle of patient autonomy into public spaces. This feat shows that a robot can maintain movement without a prepared track, even when people interrupt the rhythm and the ground changes. Consequently, we should expect more slow-and-steady roles before agile gymnasts appear, driven by falling costs.
Humanoids in Hospitals, Recycling Facilities, and Public Services
Hospitals have piloted robots for triage, delivery, and sanitation. Deployments during the pandemic are summarized throughout many research papers.
- China has showcased a smart field hospital staffed by robots, which indicates how embodied AI could extend care capacity during emergencies.
- In industry, systems with advanced visual perception that sort waste or pick mixed items are learning faster through training agents.
Functional Roles: From Waste Sorting to Public Utility
As these capabilities merge with long-duration walking, cities gain flexible helpers capable of performing several key functions:
- Carrying supplies
- Delivering lab samples
- Checking remote meters after storms
Intelligent Living Spaces in 2026 and Beyond: Designing for Humans and Humanoids
The immediate design challenge requires simple refinements to sidewalks, entries, elevators, and homes so robots can coexist efficiently without disrupting human flow. These design changes include several required elements:
- Clearer curb cuts
- Improved lighting at stairwells
- Defined access policies for charging and battery swap points
The long process of adaptation also includes cybersecurity basics, since these machines carry cameras and radios. The future of embodied AI will be shaped by recent investments in core technologies across several fields:
These advances will influence where and how embodied AI runs, from power density to edge computing and grid resilience.
Risks, Labor, and the Ethics of Walking Robots
The integration of humanoids prompts necessary questions about work, equity, and the use of public space. A specific and targeted approach is the most effective response. For instance, many jobs involve judgment, empathy, and open-ended conversation that remain hard to automate.
Evidence showing that many roles resist automation serves as a useful counterweight and a reminder that many AI systems augment humans rather than replace their vocational skills. Clear rules must be established for data use, incident reporting, and sidewalk etiquette. Furthermore, pilot programs should adopt crucial governance and audit requirements:
- Publish safety logs
- Require independent audits
When cities review the performance of walking robots on shared paths, they must include unions, disability advocates, and caregivers in the process.
The Future of Humanoid Robotics is Endurance
Agibot A2’s three-day walk signifies a critical step forward, rather than the culmination, for humanoid robotics. This achievement signals a critical shift in the field, moving the focus from laboratory novelty toward verifiable field endurance. While the distance is important, the cultural change (measuring robots by utility) is even more profound.
Embodied AI is ready to be measured by the same standards as any other urban tool (reliability, safety, energy use, and cost), as signaled by a full-size robot moving between cities under normal road rules. Expect more long-duration assignments in factories, hospitals, logistics hubs, and public spaces throughout 2026 and the subsequent years. National policies in China and other regions will push pilot programs into becoming daily routines.
Practical designs that favor slow, careful walking over spectacle will become the standard. Humanoids can support healthier aging, smoother transit, and sturdier infrastructure without turning cities into labs, provided we keep humans at the center of the design philosophy.
Frequently Asked Questions: Humanoid Deployment and Safety
Why is endurance more critical than agility for early humanoids?
Early practical assignments, such as patrols and internal logistics, prioritize continuous operation and reliability over athletic capabilities. A robot that operates for hours without fail provides immediate utility in commercial settings.
Will humanoids replace traditional wheeled robots?
No, wheeled robots remain more efficient for many jobs, particularly in flat, open environments. Humanoids are specialized for environments already built for people, offering superior navigation of stairs, doors, and standard workspaces.
What are China’s major goals for its robotics industry by 2027?
China is strategically targeting a “world-class industry” by 2027, which includes funding key components like actuators and sensors, creating shared testbeds, and integrating pilot deployments into public venues and factories.
Where will the first widespread deployments of humanoids likely occur?
Initial widespread adoption is expected in large, structured environments like hospitals, university campuses, logistics parks, and major transit hubs. Roles will focus on simple, repetitive, long-duration tasks like guidance and inventory checks.
What is the primary ethical challenge raised by walking robots?
The core ethical challenges involve defining clear rules for public safety, managing data use from onboard cameras, and addressing the impact on the labor force. Specific, transparent policies and independent audits are required.
