Topstar and Wuwen Zhike Build Industrial AI Capability Foundation, Completing a Key Piece of the Embodied Intelligence Data Loop

Topstar and Wuwen Zhike have built a physical AI data loop platform for industrial handling scenarios, covering the full chain from virtual simulation, data annotation, and data organization to model training and real-machine evaluation. The platform uses Scan2Sim and Gen2Sim to generate training scenarios for multiple operating conditions and long-tail edge cases in batches, compressing the on-site commissioning cycle from "months" to "weeks"; real data uses automatic pre-annotation plus manual review, increasing annotation efficiency 10-fold. Topstar says it has connected the full process of data collection, simulation expansion, model training, and real-machine deployment, and that all its embodied intelligence robots can rely on the platform to complete standardized iteration.