{"schema_version":"1.0","service":"Publicasta","type":"article","id":1104,"slug":"hangzhou_hangxing_no2_marathon_traffic_robot_test","title":"Hangzhou’s Hangxing No. 2 Turns a Marathon Crowd into a Public-Space Robot Test","excerpt":"Hangzhou deployed a humanoid robot around its 2026 marathon. The useful question is not whether it looks like a traffic officer, but whether a bounded patrol role can make embodied autonomy dependable in a crowded public setting.","language":"en","default_language":"en","canonical_url":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test?lang=en","image":{"url":"https://publicasta.com/storage/projects/11/pages/1104/2026/11/994a3f7e-958f-4bbc-9d29-fa3a0a79a568.webp","alt":"Humanoid traffic-support robot directing activity beside a crowded Hangzhou marathon route."},"publisher":{"id":11,"slug":"robots","name":"Robots and Autonomous Systems","url":"https://publicasta.com/robots"},"author":{"name":"Anton R"},"published_at":"2026-11-02T17:14:01+00:00","updated_at":"2026-11-02T17:14:01+00:00","content_markdown":"## What Hangxing No. 2 actually did at the marathon\n\n ![Humanoid traffic-support robot directing activity beside a crowded Hangzhou marathon route.](https://publicasta.com/storage/projects/11/pages/1104/2026/11/994a3f7e-958f-4bbc-9d29-fa3a0a79a568.webp)\n\n On Nov. 1, Hangzhou placed a humanoid traffic robot in a setting that was unusually visible, crowded, and operationally specific: the area around Huanglong Sports Center during the 2026 Hangzhou Marathon. Known as Hangxing No. 2, the machine joined a four-point patrol team assembled by Hangzhou-based DEEP Robotics and local traffic authorities. Its presence was not presented as a replacement for police officers. It was used as an additional moving unit within a larger traffic-management operation, assigned to tasks that could be defined in advance and repeated throughout the event. [The Hangzhou report](https://www.ehangzhou.gov.cn/2026-11/02/c_299511.htm) describes the robot as supporting intersection monitoring, preset-route patrols, traffic-violation warnings, and public inquiry services.\n\n That distinction matters because the deployment is easier to misunderstand when it is reduced to an image of a humanoid machine standing beside a road. The practical question was not whether Hangxing No. 2 could imitate the appearance of a traffic officer. It was whether a mobile robot could perform a bounded set of public-facing duties while people, vehicles, and runners moved through the same space. The robot was assigned to watch intersections, patrol established routes, warn about detected violations, answer questions, make standardized traffic-control gestures, and broadcast spoken safety notices. Those functions place it somewhere between a fixed sign, a public-information kiosk, and a supervised patrol assistant. They do not amount to independent command of a complex urban road network.\n\n The marathon gave the system a demanding audience without making every condition unknowable. The event drew 36,000 runners, while more than 300,000 people applied to take part. Around the sports center, that scale meant dense pedestrian movement, changing flows, and pressure on the fixed points normally used for traffic control and security. It also meant that the robot’s assignment could be tied to a particular venue, a known event schedule, designated posts, and recurring instructions. In that environment, a robot can be useful if it notices activity around an intersection, moves along a planned route, repeats a clear warning, or directs a member of the public toward the right service. The value lies in dependable repetition under supervision, not in a dramatic claim that the machine understands the entire crowd.\n\n The reported traffic role combined perception with a limited response loop. Hangxing No. 2 used its sensing and motion systems to identify situations such as pedestrians crossing the road or vehicles cutting in, then delivered warnings and performed standardized gestures while broadcasting safety advice. A person asking a question could also approach the robot for public information. These are deliberately legible interactions: the machine observes a defined type of situation, issues a recognizable prompt, and leaves the broader authority of traffic management with the human team. That arrangement makes the deployment easier for the public to understand and easier for officers to monitor than a system whose decisions are hidden behind a claim of general autonomy.\n\n The choice of a marathon also reveals the limits of what the appearance can prove. A major race compresses traffic-management work into a short period, concentrates activity at known locations, and creates many repetitive moments in which a reminder or gesture may be useful. It therefore offers a visible test of whether the robot can patrol, communicate, and remain usable around a large crowd. But it does not by itself establish that the same machine can safely resolve every ambiguous road situation, replace an officer’s judgment, or operate without intervention. In Hangzhou, the meaningful demonstration was narrower: Hangxing No. 2 became one component of an organized patrol team, helping extend attention across the event while human officers retained responsibility for the operation.\n\n That framing also connects the marathon appearance to Hangzhou’s earlier experiments with robot-assisted traffic management. The city had already deployed a 15-robot traffic-management squad at key locations, where the machines helped direct traffic, discourage violations, and provide guidance. Hangxing No. 2 brought a more explicitly humanoid platform into the same public-service conversation, but the underlying test remained practical. Could a robot move through an assigned area, deliver consistent instructions, and support officers during a crowded event? The marathon deployment offered evidence about that workflow. It did not settle the larger question of whether humanoid robots are ready to act as independent public authorities.\n\n ## The hardware behind the public-facing role\n\n Hangxing No. 2’s usefulness at the marathon depended less on its humanoid outline than on whether its body and onboard systems could support a predictable patrol. DEEP Robotics’ DR02 documentation places the platform in the 175–178 cm range, close enough to a standing adult to operate visibly among people and to present standardized gestures at roughly human eye level. That physical resemblance may make the machine legible to the public, but it is not what establishes its operational value. The more consequential question is whether the robot can move, sense, communicate, and remain serviceable while following a defined route in a busy outdoor setting.\n\n The published specifications describe a machine designed for environments where exposure and mobility matter. The DR02 is rated for IP66 protection, and its stated operating range runs from -20°C to 55°C. Those figures describe resistance and operating conditions, not a guarantee that every deployment will proceed without interruption. They nevertheless address practical constraints that a public-facing patrol cannot avoid. A traffic-support robot may encounter dust, splashing water, uneven pavement, changing temperatures, and prolonged movement. Weather resistance therefore matters more than the visual impression of a humanoid form: a robot that looks official but cannot reliably tolerate the surroundings would have little value to the officers and organizers responsible for the event.\n\n The same distinction applies to walking performance. DEEP Robotics lists a stated walking speed of 1.5 m/s. In a marathon traffic-management role, that number is not a race result and does not show that the robot can safely move at that speed through any crowd. It does indicate the intended scale of its mobility. The machine has to cover a preset patrol area, reposition between observation points, and remain useful without being treated as a stationary display. Route mobility is the relevant capability here. The task is not to cross a finish line or imitate an officer theatrically; it is to maintain a presence at designated locations and support a workflow that already has human supervision.\n\n Power management is similarly more important than appearance. The documentation identifies a swappable 1,440 Wh battery. A replaceable battery gives operators a way to manage the practical gap between a robot’s operating cycle and the duration of an event. It does not, by itself, establish how long Hangxing No. 2 remained active, how often it required a change, or how much labor was involved in preparing and replacing the battery. Those details would be needed to assess endurance in real deployments. Still, the design choice points to an operational priority: public-service robots must be maintainable in the field, not merely capable of performing a short demonstration under controlled conditions.\n\n The sensing package is what turns the body into a mobile observation platform. The supplied specifications identify LiDAR, depth sensing, and wide-angle cameras among the robot’s sensors. These systems serve different but related purposes. LiDAR can contribute to spatial mapping and obstacle detection; depth sensing helps estimate the position of nearby objects; wide-angle visual input broadens the machine’s view of people, vehicles, and the surrounding route. Together, they provide the raw information needed for navigation and situational awareness. They do not automatically give the robot a reliable understanding of every human action. A sensor can detect a person or obstacle without resolving whether a gesture is a request for help, a warning to stop, or an accidental movement in a dense crowd.\n\n The reported 275 TOPS compute unit is another specification that should be read as system capacity rather than proof of autonomy. More onboard computing can support the processing of multiple sensor streams and the algorithms needed for perception, navigation, and interaction. It does not reveal how often the system misreads a scene, how it behaves when sensor inputs conflict, or when a human operator must intervene. Those outcomes depend on software, route design, communication links, and operating procedures as much as on the headline compute figure.\n\n That is why the marathon report’s description of patrols, warnings, gestures, and spoken safety notices matters alongside the product sheet. The event showed a robot assigned to a bounded public-facing role, while the specifications describe the equipment that can make such a role possible. Neither source establishes that the robot can independently manage ordinary city traffic or replace police judgment. The hardware supports a narrower claim: with suitable sensing, protected construction, mobile power, and route-capable movement, a humanoid platform can be placed where people need to see and hear it. Whether that becomes dependable public service depends on the operating discipline around the machine—its route, supervision, maintenance, and response when the scene no longer matches the conditions it was designed to handle.\n\n ## Why a marathon is a useful but narrow autonomy test\n\n A marathon offers a demanding-looking environment without being an entirely open-ended one. The crowd can be dense, movement can be difficult to predict, and intersections require attention, but the event itself is bounded by time, place, and procedure. The robot is not being asked to understand an entire city at once. It is assigned to known locations, follows preset patrol routes, and operates within an event whose traffic arrangements have already been organized by human authorities. That structure matters: it turns a broad claim about “autonomous” public-service robots into a more specific question about whether a machine can perform a limited set of duties reliably amid heavy foot traffic.\n\n The repetitive nature of the work also makes the setting useful for evaluation. Safety notices, standardized gestures, traffic-violation warnings, and answers to routine public questions are not identical tasks, but they are comparatively legible compared with the unpredictable decisions faced by a traffic officer in ordinary city conditions. A robot can be assessed on whether it reaches the required points, repeats instructions consistently, recognizes when people are blocking or crossing a controlled area, and communicates in a form that bystanders understand. The presence of nearby human police further narrows the robot’s responsibility. It can support the operation while leaving consequential decisions and physical intervention with people who have the authority to make them.\n\n That arrangement is also the reason the marathon should not be treated as proof that a humanoid robot is ready to replace officers. A controlled event can reduce the range of situations the system must handle. Roads and patrol areas are known in advance; instructions are repeated; the deployment window is limited; and the human team can intervene when the robot reaches the edge of its competence. Those conditions are valuable because they make a first public test safer and easier to observe, but they also conceal the harder parts of everyday traffic management. City streets do not reliably provide a preset route, a predictable crowd, or an obvious division between normal behavior and an emergency.\n\n The unresolved cases begin with perception. A dense crowd can hide a person, vehicle, sign, or hand signal from the robot’s sensors. An apparently simple gesture may be ambiguous, especially when several people are moving at once or when pedestrians do not know what the machine is asking them to do. The question is not merely whether the robot can issue a warning, but whether it can determine when a warning is appropriate, distinguish compliance from hesitation, and avoid escalating an ordinary misunderstanding. A system that responds safely to clear, repeated patterns may still need human judgment when people behave in ways that are unusual but not dangerous.\n\n There are practical questions as well. If the robot encounters an obstruction, loses its route, or receives conflicting signals from the crowd, can it stop in a safe place and request assistance? Who takes responsibility when its warning is wrong, its patrol is interrupted, or an incident develops faster than a human supervisor can respond? A public deployment also raises questions about the handling of sensor data and the boundaries of observation in a crowded space. Battery changes and other maintenance needs matter because a patrol is useful only while the robot can remain available, and a system that requires frequent intervention may shift work to its human support team rather than remove it.\n\n For that reason, the marathon is best understood as a narrow autonomy test: meaningful because it places embodied software and hardware in a real public setting, narrow because the setting is deliberately structured. Its strongest result would not be a dramatic demonstration of independent intelligence. It would be evidence that the robot can perform a defined patrol role, communicate consistently, remain within safe operating limits, and hand difficult cases back to human officers. The next question is whether those conditions can be measured and repeated when the crowd is less predictable, the route is less controlled, and the consequences of a mistake are harder to contain.\n\n ## The practical benchmark is coordination, not theatrical autonomy\n\n Hangzhou’s marathon appearance matters less as a demonstration that a humanoid robot can stand beside a roadway than as a test of whether a machine can fit into an existing public-safety operation. The city had already described a 15-robot traffic-management squad earlier in 2026, presenting the machines as part of a wider effort to use robots in traffic control and event support. The marathon therefore raises a more useful question than whether Hangxing No. 2 looked convincing in a traffic-police role: could it perform a limited assignment while human officers retained authority over the situation?\n\n That distinction matters because public events reward coordination. A robot can be useful without independently understanding every person, vehicle, or emergency around it. It can occupy a known point, follow a preset route, repeat a standard instruction, signal with a defined gesture, and pass unusual situations to people responsible for decisions. The marathon setting—large, scheduled, concentrated around known locations, and supported by police—makes that division of labor possible. Hangzhou has also presented event technology as a broader system, including drone-assisted emergency support. In that context, the robot is best understood as one interface in an operating network, not as a substitute for the network’s human judgment.\n\n Future deployments should be judged against operational measures rather than spectacle. The first is uptime: how long the unit remained available, and how often it had to be paused, recharged, reset, or removed from duty. A second is intervention rate: how frequently a human officer had to take over because the robot encountered an ambiguous movement, an obstructed route, an unsafe interaction, or a task outside its programmed boundaries. Those figures would reveal more about practical autonomy than a video of a robot issuing a warning at a clear intersection.\n\n False warnings deserve equal attention. If the system repeatedly warns people who are not violating an instruction, the burden shifts back to officers and the public may learn to ignore it. If it is too cautious, it may fail to address the situations for which it was deployed. A credible report would separate warnings issued, warnings judged appropriate, and incidents that required correction. Route coverage should be measured in the same spirit: not merely the distance assigned to the robot, but the proportion of that route it completed under the required conditions, including pauses, blocked paths, crowd spillover, and handoffs to human staff.\n\n Incident handling is another decisive test. A public-service robot does not need to solve every emergency, but it does need a clear response when normal operation breaks down. Future accounts should say what happened when a person asked a question the system could not answer, when a gesture was unclear, when pedestrians obstructed movement, or when an officer needed the robot to stop immediately. The key issue is not whether the machine can avoid every exception. It is whether the exception is detected quickly, escalated to the right person, and recorded in a way that supports review.\n\n The economics are similarly practical. Maintenance cost should include more than the purchase price: battery handling, charging or swapping, transport, software support, inspection, cleaning, repairs, and the staff time needed to supervise the unit all affect whether a deployment can be repeated. The DR02 documentation’s swappable battery design is relevant to that question, but a specification alone does not show how many battery changes an event requires or how much labor they involve. Those are the figures that would let organizers compare robotic support with other ways of extending traffic-management capacity.\n\n Human authority must remain explicit. Officers should be able to pause the system, redirect it, or remove it from the public space without negotiating with the machine. The public should also be able to tell when a robot is giving a standardized notice and where responsibility lies if that notice is misunderstood. A humanoid form may make the unit more legible at a glance, but legibility is not accountability. The chain of command, escalation rules, and records of intervention matter more than the appearance of autonomy.\n\n That is the useful benchmark for Hangxing No. 2 and for the larger squad around it. If Hangzhou publishes measured uptime, intervention rates, false-warning rates, route coverage, incident outcomes, maintenance costs, and the limits of human authority, the marathon can become evidence of a repeatable public-service workflow. Without those measures, it remains mainly a persuasive scene: a robot in a police-like role, surrounded by a crowd, demonstrating that the technology can be displayed. The meaningful advance would be narrower and more valuable—the ability to integrate embodied machines into bounded civic tasks without obscuring who remains responsible for the result.","available_translations":[{"language":"ar","title":"روبوت هانغشينغ رقم 2 من هانغتشو يحوّل حشود الماراثون إلى اختبار للروبوتات في الأماكن العامة","html_url":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test?lang=ar","markdown_url":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test.md?lang=ar","json_url":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test.json?lang=ar","api_url":"https://publicasta.com/api/public/v1/channels/robots/articles/hangzhou_hangxing_no2_marathon_traffic_robot_test?lang=ar"},{"language":"de","title":"Hangzhous Hangxing Nr. 2 macht die Marathon-Zuschauermenge zum Testfeld für einen Roboter im öffentlichen Raum","html_url":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test?lang=de","markdown_url":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test.md?lang=de","json_url":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test.json?lang=de","api_url":"https://publicasta.com/api/public/v1/channels/robots/articles/hangzhou_hangxing_no2_marathon_traffic_robot_test?lang=de"},{"language":"en","title":"Hangzhou’s Hangxing No. 2 Turns a Marathon Crowd into a Public-Space Robot Test","html_url":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test?lang=en","markdown_url":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test.md?lang=en","json_url":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test.json?lang=en","api_url":"https://publicasta.com/api/public/v1/channels/robots/articles/hangzhou_hangxing_no2_marathon_traffic_robot_test?lang=en"},{"language":"es","title":"Hangxing No. 2 de Hangzhou convierte el público de un maratón en una prueba para robots en espacios públicos","html_url":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test?lang=es","markdown_url":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test.md?lang=es","json_url":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test.json?lang=es","api_url":"https://publicasta.com/api/public/v1/channels/robots/articles/hangzhou_hangxing_no2_marathon_traffic_robot_test?lang=es"},{"language":"fr","title":"Le Hangxing n° 2 de Hangzhou transforme la foule d’un marathon en test grandeur nature pour un robot dans l’espace public","html_url":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test?lang=fr","markdown_url":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test.md?lang=fr","json_url":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test.json?lang=fr","api_url":"https://publicasta.com/api/public/v1/channels/robots/articles/hangzhou_hangxing_no2_marathon_traffic_robot_test?lang=fr"},{"language":"pl","title":"Hangzhou testuje Hangxinga nr 2: tłum maratończyków jako sprawdzian robota w przestrzeni publicznej","html_url":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test?lang=pl","markdown_url":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test.md?lang=pl","json_url":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test.json?lang=pl","api_url":"https://publicasta.com/api/public/v1/channels/robots/articles/hangzhou_hangxing_no2_marathon_traffic_robot_test?lang=pl"},{"language":"ru","title":"Hangxing № 2 из Ханчжоу превратил марафонскую толпу в испытание робота в общественном пространстве","html_url":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test?lang=ru","markdown_url":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test.md?lang=ru","json_url":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test.json?lang=ru","api_url":"https://publicasta.com/api/public/v1/channels/robots/articles/hangzhou_hangxing_no2_marathon_traffic_robot_test?lang=ru"},{"language":"zh","title":"杭州“杭行二号”将马拉松人流变成公共空间机器人测试场","html_url":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test?lang=zh","markdown_url":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test.md?lang=zh","json_url":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test.json?lang=zh","api_url":"https://publicasta.com/api/public/v1/channels/robots/articles/hangzhou_hangxing_no2_marathon_traffic_robot_test?lang=zh"}],"_links":{"self":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test.json?lang=en","api":"https://publicasta.com/api/public/v1/channels/robots/articles/hangzhou_hangxing_no2_marathon_traffic_robot_test?lang=en","html":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test?lang=en","canonical":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test?lang=en","markdown":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test.md?lang=en","json":"https://publicasta.com/robots/hangzhou_hangxing_no2_marathon_traffic_robot_test.json?lang=en","channel":"https://publicasta.com/api/public/v1/channels/robots","channel_articles":"https://publicasta.com/api/public/v1/channels/robots/articles","search":"https://publicasta.com/api/public/v1/search","documentation":"https://publicasta.com/api-docs#reading-publicasta","openapi":"https://publicasta.com/api-docs/openapi.json","llms":"https://publicasta.com/llms.txt"}}