Abstract:
1.Military special robots cater to the development trend of unmanned and intelligent warfare in the future.
Military robots, such as quadruped robotic dogs and robotic wolves, can perform tasks such as reconnaissance, patrolling, material transport, and combat support in complex battlefield environments, effectively reducing personnel casualty risks and improving operational efficiency. At the same time, these robots possess autonomous decision-making and multi-scenario adaptability, allowing them to replace humans in high-risk areas and become an important part of the future intelligent warfare system.
In the United States: Boston Dynamics, Palantir Technologies, Open AI, and other well-known companies have announced collaborations with the U.S. military. In China: Since last year, China’s military industry has launched military robots. The Chinese military robots adopt a model of state-owned enterprises + private enterprises, with the “robotic wolf” leading new demands for military robots.
Auxiliary functions“robotic dogs“ made their debut during the joint military exercise in June 2024, where they served as companions to soldiers on the battlefield, assisting in reconnaissance and preemptive strikes in specific environments. At the 15th Zhuhai Airshow in 2024, the quadruped “robotic wolf” developed independently by the China Ordnance Industry Group made its first appearance, attracting widespread attention. The change in name from “dog” to “wolf” signifies a clear distinction from the previous “robotic dog” and highlights the upgraded combat methods and enhanced capabilities of the new quadruped bionic robots..
“The robotic wolf” adopts a bionic design, simulating the movement patterns of real wolves, and possesses high flexibility and strong adaptability. Through advanced sensors and algorithms, it can maintain balance and stability in complex terrains (such as mountains and deserts). These “robotic wolves” are highly intelligent, equipped with artificial intelligence algorithms that enable autonomous navigation, obstacle avoidance, and target recognition, allowing them to complete tasks without human intervention. Furthermore, “robotic wolves” can carry lightweight weaponry, such as automatic rifles and grenade launchers, to execute combat missions. The emergence of the “robotic wolf” marks a solid step forward for China in the field of unmanned combat, providing new guarantees for national defense security.
2. The future development direction of artificial intelligence is embodied intelligent AI robots.
2024 has witnessed the rapid rise of several startups in this field, with notable representatives including Qianxun Intelligent, Xinghai Map, WorldLabs, and Noematrix. These advancements indicate that embodied intelligence will permeate various aspects of human life in the coming years.
➢ Qianxun Intelligent: Within six months of its establishment, Qianxun Intelligent completed two rounds of financing, focusing on building highly generalized and versatile humanoid robots, utilizing visual language models (ViLa) and component constraint models (CoPa) to solve data acquisition challenges, and plans to apply its products in commercial, service, and household scenarios.
➢ Galaxy General: Galaxy General has developed the first end-to-end navigation large model based on single-view video streams from robots. This large model does not require mapping or signals from sensors other than RGB cameras, and it has first validated the feasibility of the first pure visual “FSD” model in the field of embodied intelligence. The action data of this model comes entirely from simulated synthesis of the real world, achieving zero-shot cross-scenario generalization capability in the real world without using any real-world action data.
➢ Xinghai Map: Proposed the concept of “one brain, multiple forms,” with its embodied foundational model EFM and spatial intelligence engine RSR achieving the ability to learn complex tasks from a small number of demonstrations, successfully launching the full-size wheeled dual-arm embodied body R1.
➢ World Labs: Co-founded by “AI Mother” Fei-Fei Li, emphasizing the importance of spatial intelligence. World Labs’ first “world model (creating the world)” was officially released at midnight on December 4, which represents the current stage of spatial intelligence technology. It can generate detailed 3D scenes from a single image and supports real-time rendering and controllable camera effects. Users only need to provide a 2D image to automatically generate a detailed 3D scene.
➢ Noematrix: Focuses on force control intelligent technology, enabling robots to perform precise tasks in physical environments. Force control intelligence emphasizes the perception, reasoning, and interaction capabilities of AI systems in the physical world, with a particular focus on the application of mechanical sensors and control algorithms. Professor Lu Ce Wu pointed out that true force control intelligence requires not only strong algorithmic support but also a deep understanding of hardware to achieve efficient operation in the physical world.
➢ ByteDance: The former head of the “Tongyi Qianwen” large model technology at Alibaba, Zhou Chang, left and subsequently led his team to join ByteDance. According to news reports, what Zhou is preparing to do at ByteDance may be similar to Fei-Fei Li’s “large world model,” generating a 3D world from a single image.
3. Investment Recommendations: In the field of embodied intelligence + AI, there is potential to empower various product forms of robots through large models. Currently, most algorithm training is still based on visual algorithms + AI large models, focusing on companies with positioning attributes in the visual algorithm field such asOrbbec.
The robotics + military field is expected to become a new emerging application scenario, with related military robot companies such asJianshe Industrial, Lingyun Co., and Donghua Testing;
Other civilian quadruped robot fields, as well as companies benefiting from advancements in AI large models such as Zhongjian Technology.
4. Risk Warning: Progress in embodied intelligence-related model algorithms may fall short of expectations, and demand for military and civilian robots may not meet expectations.
Source: Minsheng Securities, various company websites, and other public information, for sharing purposes only, does not represent my personal stance. If you believe that the platform’s article infringes your intellectual property rights, please contact us promptly, and we will delete it as soon as possible.Source: Aiden’s Hard Technology Research