Core Insights - The "Junpu Intelligent-Zhiyuan Joint R&D Laboratory" has announced the operational testing of the new generation wheeled robot G2, marking a significant step in advancing artificial intelligence from "perceptual intelligence" to "executive intelligence" in industrial flexible manufacturing [1][2] Group 1: Laboratory and Development Focus - The joint laboratory aims to create a versatile intelligent toolchain system focused on core process units in industrial scenarios [1] - Current research emphasizes system framework construction, core algorithm optimization, and equipment collaboration capabilities to establish a stable and reliable underlying support system for industrial intelligent devices [1] Group 2: Algorithm and Performance Metrics - The laboratory has adopted a "human-machine loop-sample efficient reinforcement learning" paradigm, achieving micrometer-level contact flexible assembly with a robot system success rate of approximately 99% and a reinforcement learning placement success rate of 99.33% [1] - The visual grasping success rate stands at 100%, showcasing the effectiveness of the developed algorithms [1] Group 3: Robot G2 Capabilities - The G2 robot has achieved integrated "hand-eye-body-network" collaboration, providing replicable and scalable intelligent solutions for the flexible upgrade and high-quality development of the manufacturing industry [2] - G2 operates with an average task cycle of 15.28 seconds, with the fastest cycle reaching 12.97 seconds, and supports mixed deployment of multiple machines and data sharing [1]

Core Insights - Ningbo Junpu Intelligent Manufacturing Co., Ltd. has established a joint research laboratory with Zhiyuan Robotics, focusing on advancing industrial AI tools and algorithms [1][2] - The laboratory aims to transition AI from "perceptual intelligence" to "executive intelligence," creating a closed-loop ecosystem for the development and validation of tools [1] - The initiative addresses challenges in acquiring high-quality industrial private data by building a complete closed loop from scene setup to data collection, model training, and tool development [1] Group 1 - The joint laboratory will leverage the new generation wheeled robot G2 for continuous training and optimization in highly realistic simulated environments [1] - A physical data collection center is being established to gather critical data from real and simulated scenarios, ensuring that the tools and models developed are practical and continuously iterated [1] - The laboratory has already formed partnerships with industry players to promote the industrial application of AI vision models [1] Group 2 - The joint laboratory will strategically collaborate with the Ningbo Embodied Intelligent Robot Innovation Center, which serves as an open platform for industrial demands and data [2] - Tools and data platforms developed by the joint laboratory will be made accessible to customers and partners through the innovation center, enabling them to autonomously extract data and utilize models [2] - This collaboration aims to create a "technology-business dual closed loop" from technology development to commercial implementation, ensuring sustainable growth [2]