📰 ZhiYuan has released the AGILE 2.0 integrated sensing-and-control model. The focus isn’t simply to make robots “see,” but to combine environmental perception, terrain understanding, whole-body motion control, and operation into a single system.
🔥 The model uses a vision end-to-end approach. The Lingxi X2 it runs on has already completed dynamic tasks such as ball-kicking walking, jumping rope, and collaboratively carrying boxes. Honestly, these tests don’t look easy— the robot has to judge the environment while immediately adjusting its own movements.
💡 In the past, when watching robot demos, many times the robots followed fixed routes and pre-prepared actions. I didn’t expect this time to emphasize real-time perception during motion and making adjustments based on changes. This direction is clearly closer to the requirements of real-world use.
In fact, ball-kicking walking tests balance and continuous control; jumping rope requires coordinated motion; and collaboratively carrying boxes is more like humans and robots working together to complete a task. Putting these three types of tasks together shows that AGILE 2.0 isn’t validating a single action—it’s checking whether perception and control can work together.
🤔 Once robots truly enter factories, warehouses, and even everyday environments, they won’t face perfectly uniform floors or fixed instructions. Can the dynamic tests of the Lingxi X2 become a practical step for the integrated sensing-and-control model?
#智元机器人 #AGILE2.0 #人形机器人 #ArtificialIntelligence
🔥 The model uses a vision end-to-end approach. The Lingxi X2 it runs on has already completed dynamic tasks such as ball-kicking walking, jumping rope, and collaboratively carrying boxes. Honestly, these tests don’t look easy— the robot has to judge the environment while immediately adjusting its own movements.
💡 In the past, when watching robot demos, many times the robots followed fixed routes and pre-prepared actions. I didn’t expect this time to emphasize real-time perception during motion and making adjustments based on changes. This direction is clearly closer to the requirements of real-world use.
In fact, ball-kicking walking tests balance and continuous control; jumping rope requires coordinated motion; and collaboratively carrying boxes is more like humans and robots working together to complete a task. Putting these three types of tasks together shows that AGILE 2.0 isn’t validating a single action—it’s checking whether perception and control can work together.
🤔 Once robots truly enter factories, warehouses, and even everyday environments, they won’t face perfectly uniform floors or fixed instructions. Can the dynamic tests of the Lingxi X2 become a practical step for the integrated sensing-and-control model?
#智元机器人 #AGILE2.0 #人形机器人 #ArtificialIntelligence
