Minglue Tech and Hikrobot Join Forces at WRC 2026, Targeting Commercial Robotics with 'Agent + Embodied' Approach

According to a keynote transcript published by Leiphone.com, Wu said that robots in the physical world mostly provide 'emotional value' today, and true productivity requires overcoming the last mile of engineering delivery, business process reengineering and on-site deployment. He also revealed that Minglue is collaborating with Hikrobot, leveraging its own strengths in software, agents and models while Hikrobot contributes mobile robot R&D, motion control, sensor fusion and system integration.

On the exhibition floor, the joint booth demonstrated three scenarios: restaurant cleaning, industrial logistics and intelligent patrol. In the restaurant scenario, a wheeled humanoid robot and a wheeled dual-arm robot worked together to clear tables, collect dishes and scrub floors. After large-scale training, the robots could distinguish kitchen waste from personal belongings such as phones and keys, removing only the trash. In logistics, robots showed a full chain of flexible picking from high-rack retrieval to fine sorting, while patrol robots demonstrated autonomous navigation and anomaly detection.

Wu introduced the concept of two 'brains' for the second half of robotics. The first is the conventional robot brain based on VLM/VLA models for reasoning and action planning. The second is an organizational brain that links robots to each other, to agents, and to existing IT systems. Minglue’s open-source human-machine collaboration platform, Octo, is designed to form such an organizational brain. According to Wu, Octo will eventually connect both digital-world agents and physical-world robots, possibly via open protocols.

Minglue has long explored the HAO intelligent framework (Human, AI, Organization), proposed in 2018. The company believes intelligent labor in the AI era takes two forms: agents in the digital world and robots in the physical world. Minglue aims to integrate both into organization networks with shared memory. Wu drew parallels between Agentic AI and embodied intelligence, noting that a similar trajectory from rigid rules to end-to-end models and structured skills can be observed. He cited examples like atomic skill libraries and memory systems, which are equally important for robots.

Wu also outlined a five-stage evolution path from L1 to L5 for AI-native organizations, including token readiness, work online, structured memory, complex collaboration networks and continuous improvement. He stressed three principles for deploying robots: redivision of labor, process reengineering, and being human-friendly. He cited an example of a dishwashing robot that failed to improve a restaurant’s efficiency because the human worker performed other tasks as well, underscoring the need for process redesign. He also mentioned the role of Forward Deployed Engineers (FDEs) in embedding robots into real workflows.