In the first quarter of this year, my country’s industrial robot output increased by 33.2% year-on-year.Behind the increase, domestic robots are answering three questions: Can the algorithm be smarter?Can you make core components yourself?Can the application scenario be penetrated deeply enough?
In the theater, the popcorn robot of Shenzhen Yuejiang Technology Co., Ltd. is "at work": it visually senses the position of the cup, plans the grabbing path, scoops up the popcorn with its robotic arm, accurately fills the cup and hands it to the customer, without any human intervention during the whole process.
The action may seem simple, but actually involves core embodied intelligence technologies such as multi-modal perception, real-time decision-making, high-precision force control and dexterous operation."We have built a 'brain + cerebellum' autonomous evolution system. The brain is a self-developed 'Kongyi DobotWAM' embodied large model, and the cerebellum is a high-precision whole-body force control system." Xie Kaixuan, marketing director of Yuejiang Technology, said, "In the face of complex multi-step tasks, the large model first""Rehearse in your head" and then execute. In this way, "think first and then act" can reduce trial and error costs and safety risks.The popcorn robot is a real commercial scenario and can complete 1,000 cups a day without making mistakes."
On the other side, the cleaning robot of Shenzhen Pudu Technology Co., Ltd. has just "finished work."It only takes action when it "sees" garbage, and is on standby the rest of the time.According to the staff, the reason why the cleaning robot can "deploy on demand" relies on the self-developed visual AI model, which can accurately identify and dynamically plan the path.
The technical architecture built by Pudu is called "One Brain, Many Shapes", which means it has self-developed a set of low-level general algorithm models that can simultaneously drive specialized, humanoid, humanoid, quadruped and other forms of robots."The technical base is the same. We only need to do adaptation training based on different operating scenarios and task requirements, and it can be implemented quickly." Zhang Tao, founder and CEO of Purdue, said that "one brain, multiple shapes" has greatly reduced research and development costs, and the product launch cycle has also been significantly shortened.
Domestic robots no longer complete actions through a single remote operation or preset programming, but perform tasks completely autonomously in an open and uncertain real-life environment.Behind this is the fact that robot companies start from the bottom-level algorithms so that the robot's decision-making capabilities can truly adapt to complex production line environments and life scenarios.At the same time, robotics companies are accelerating their independent research on core components and parts, gradually easing the "stuck neck" problem that has plagued the industry for many years.
In the workshop of a precision manufacturing giant, the repetitive positioning accuracy requirement for loading and unloading robots reaches an extremely demanding ±0.02 mm.Cross-border robots not only meet the standards, but also achieve large-scale deployment of more than 300 units, relying on self-developed high-performance control systems and motion algorithms."We insist on self-research on the whole stack." Xie Kaixuan said that independent controllability is not only the source of cost advantages, but also the fundamental guarantee for rapid iteration, ensuring supply chain security, and migrating industrial-grade reliability to downstream scenarios.
“By increasing R&D investment, reliability verification, industry-university-research collaboration, etc., the company continues to narrow the gap with top foreign components.” Wang Zhicheng, general manager of the Product R&D Center of Guangdong Topstar Technology Co., Ltd., told reporters that Topstar has independently mastered the three core component technologies of industrial robots: controllers, servo drives, and perception systems.In the first quarter of this year, the company's industrial robot and automation application system business revenue increased by 81.20% year-on-year.
The significance of these breakthroughs is not only to reduce the cost of a robot, but also to strengthen the safety base of the entire industry.Another advantage of the development of my country's robot industry is that it has rich industrial scenarios and massive application scenarios, which is the best "training ground" for robots.
"Robotics technology must "grow" from real scene needs." Xie Kaixuan said, "my country's industrial scenes are very diverse. Cross-border robots have covered more than 15 industries and 200 subdivided scenes, and have accumulated massive process data. By continuously accumulating operating experience and optimizing strategies, robots are getting smarter."
The actual problems encountered by each robot on the production line will become nourishment for upgrades."Robotics companies want to be "hexagonal warriors." Technology, products, supply chain, commercialization, strategy, and organization must not be short on any one of them," Zhang Tao said.
