Hao Zheng

For home service robots to work reliably in complex, dynamic and highly unstructured homes, the mechanical body, perception, motion control, task planning, AI models, data feedback loops and engineering platform must work together. From joints, end effectors, sensors and force control to navigation, real-time control, task scheduling and cloud services, every part affects whether a robot can complete its task.

In real homes, communication latency, motion accuracy, hardware-software integration, data quality, model generalization, long-horizon task stability, recovery from failures and remote operations all become challenges at once. Capabilities that work in the lab need continuous testing, system tuning and engineering iteration to adapt to different layouts, furniture, objects and habits. Drawing on these practical problems and lessons, this talk traces the full stack from hardware to intelligence and the work required to turn algorithms into stable, maintainable products that can keep improving.