Q: Why is Tianchuang making humanoid robots?
A: As a technology company serving industries such as energy and municipal engineering with robotics and AI algorithms as its core products, our vision has always been to introduce cutting-edge robotics technology into the field of industrial intelligent operation and maintenance. Our Tiankui series of industrial explosion-proof heavy-duty humanoid robots are more accurately described as robots capable of operating in hazardous industrial environments. They represent the next level of our existing industrial inspection robots and form a crucial foundation for building a closed loop of capabilities from problem discovery to problem resolution in the field of industrial intelligent operation and maintenance.
Q: What are the differences between Tiankui explosion-proof humanoid robots and mainstream general-purpose humanoid robots?
A: We've been thinking about what kind of humanoid robots our customers would need, given that we're a B2B robotics company focused on real-world application. The key difference between industrial and residential robotics lies in the complexity of the working environment and the high precision required for operations. For us, the first priority is to improve the robot's protective capabilities. This is Tianchuang's strength; for eight years, we've been dealing with complex and dangerous industrial scenarios, accumulating extensive experience in structural design to ensure robots can reliably survive in environments with high temperature and humidity, strong magnetic fields, dust, and flammable/explosive conditions. Therefore, our first goal was to design a humanoid robot with a high level of protection, even explosion-proof capabilities. We've achieved this; our Tiankui series robots are likely the world's first humanoid robots to receive IICT Level 6 explosion-proof certification. We've implemented numerous special treatments in areas such as motion structure, power system, and electrical wiring. The second point is high-precision operation capabilities, which presents a significant challenge. This isn't just a structural issue; more importantly, it involves advanced AI applications such as multimodal perception, force feedback and control, and the identification and fault diagnosis of industrial equipment. Current large-scale models cannot fully support these types of operational applications. Therefore, we plan to address this issue in several steps. The first step is to define the industrial embodied robot. We continue the biomimetic logic of humanoid robots, also employing dual-arm operation, but with enhanced end-effector load. In its fully extended horizontal configuration, a single arm can carry up to 15kg of load, matching the high torque and high load requirements of industrial scenarios. We've also optimized the waist structure; this generation features a folding and lifting waist, allowing the arm to extend vertically to a maximum end-effector height of 3m, significantly improving the efficiency of frequent high-level operations in industrial settings. Furthermore, the chassis utilizes our expertise in wheeled structures, resulting in higher robot mobility and longer battery life. The second step involves a combination of remote operation and pre-training to address the current challenges of operational capability in industrial settings. We've designed a cockpit that allows for rapid navigation to the work site. Users can then operate the robot using a force-feedback remote controller within the cockpit. These operations are recorded throughout, serving as samples for continuous training of the robot model. For tasks with clearly defined environments, objects, and tasks, such as switching on/off, valve turning, and drilling, pre-training has been conducted, allowing back-end personnel to complete these tasks with a single click from the control panel. In the future, we will continuously train the robot's perception, motion control, and task decision-making capabilities through on-site human-robot collaborative operations, gradually achieving fully embodied intelligent operation.
Q: Why does Tiankui's explosion-proof humanoid robot use a wheeled chassis instead of a legged structure?
A: Both legged and wheeled systems have their advantages and disadvantages, but we focus more on scenario-based thinking. In the industrial sites we serve, wheeled systems can cover more than 90% of the work scenarios, and even for those requiring multi-level access, there are usually industrial elevators on site. From the user's perspective, accessibility is more about reliability and economy; users are more concerned with the operational capability itself. Therefore, from the beginning, we focused our R&D efforts on the upper body, emphasizing the operational and load-bearing capabilities of the booms. The chassis uses our common four-wheel, eight-wheel drive, explosion-proof wheeled architecture, which has long endurance, multi-terrain adaptability, and multi-sensor fusion navigation capabilities, fully meeting our accessibility needs in complex industrial environments. Legged systems still face some challenges in terms of walking stability, endurance, and explosion-proof structural design, and may be more suitable for lightweight consumer-end scenarios.
Q: In which industrial scenarios can Tiankui explosion-proof humanoid robots be applied?
A: Our Tiankui robots can be used not only for the opening and closing of many local switches and the operation of valves in the power industry, such as switching and operation of circuit breakers in power grid substations, but also for the remote operation of transmission and transformation equipment in offshore wind power substations. They can also be used for complex remote operation maintenance in scenarios such as coal transportation, water treatment, and boilers in thermal power plants, and even for maintenance work in the micro-radiation environment of nuclear power plant nuclear islands. Furthermore, they can be used for valve switching and motor starting/stopping in explosion-proof scenarios in the petrochemical industry, reducing the safety risks for on-site personnel. Theoretically, our Tiankui humanoid robots can replace humans in complex industrial scenarios, especially hazardous ones, and even in future maintenance. In some emergency rescue scenarios, such as bomb disposal and demolition, they can also be used to replace humans through remote operation.
Q: What are Tianchuang's views on the future development of humanoid robots?
A: The application of humanoid robots is highly certain, but the road ahead will still be full of twists and turns. Humanoid robots aren't actually a structural innovation; they're a culmination of existing technologies. It's just that the rapid development of AI technology in recent years has brought more possibilities to the application of humanoid robots. The integration of large-scale models and humanoid robots can greatly improve the multimodal perception and recognition capabilities and interaction capabilities of humanoid robots, and even improve their motion planning and control capabilities end-to-end. However, frankly speaking, there's still a long way to go before they can be universally applied. For example, any small human movement in the real physical world requires extensive pre-training if a humanoid robot needs to imitate it and achieve generalization in different scenarios. For humanoid robots to achieve embodied intelligence, it's not just about finding scaling laws like with large-scale models; there's still a long way to go even in terms of the general structure itself. For example, our current joint motors are far from achieving the load-bearing ratio of human muscles, which greatly limits the robot's operational capabilities.
Another issue is the commercialization of humanoid robots. Currently, most humanoid robots are positioned for consumer-facing (to C) applications, such as domestic services and elderly care. This presents an inherent contradiction: while to C scenarios offer some tolerance for error, they demand higher generalization capabilities because these scenarios are less standardized and more uncertain. There's also the cost issue. Even if the final price of humanoid robots drops below 100,000 yuan in the future, it's still too high for individual consumers, limiting the large-scale commercial application of humanoid robots. Business-facing (B2B) applications, on the other hand, have stronger purchasing power, especially in complex and dangerous scenarios. Their replacement of humans offers not only cost savings but also added safety value, which is the core reason why B2B clients are willing to pay a premium. Furthermore, B2B scenarios have stronger standardization and regulation. Although the tolerance for error is lower, this makes rule-based AI training in B2B vertical scenarios technically more feasible.
Therefore, from Tianchuang's perspective, humanoid robot applications in the B-end are worth investing in and persisting with, and may even achieve large-scale implementation earlier than in the C-end.
