Embodied AI on the Waste Sorting Line: DataBeyond’s PICKINGMAN in Industrial Practice

The industrial value of embodied AI needs to be demonstrated in actual production. The tasks robots can perform, their ability to handle challenging conditions, and their integration into production lines are key considerations for practical deployment. For DataBeyond, waste sorting is an application with clear demand that merits sustained development.

On October 3, People’s Daily reported on Shangyu, Zhejiang, and its approach to developing the embodied AI industry. Starting with demand for chemical plant inspections, the district connected practical applications with research and manufacturing. This approach offers a useful lesson: specific industrial needs can guide technology development, while actual deployment provides a basis for validating and refining products. Source

Waste sorting requires the same focus on practical needs. Plastic bottles arriving at a processing facility may be deformed, food containers may be contaminated, and different materials may overlap or obscure one another. Their shapes and positions vary constantly. Sorting must keep pace with the production line, while equipment must withstand dust, moisture and dirty conditions.

These conditions require robots to connect perception, decision-making and execution: identifying a target, determining how to grasp it, and moving it into the appropriate recovery stream. Product development must address these specific tasks, with each design choice responding to an operational requirement.

DataBeyond’s PICKINGMAN embodied AI robot was developed specifically for mixed municipal solid waste sorting. According to published product information, PICKINGMAN has two arms with a combined 14 degrees of freedom, coordination between high-level decision-making and motion control, and an IP65 dust and water protection rating. It can sort PET bottles and trays, PP food containers and milk tea cups, PE materials, aluminium cans, glass and other mixed plastics. Product information

DataBeyond has long focused on developing technologies for recyclable material sorting and providing integrated sorting solutions for entire facilities. Its understanding of materials, processes and customer requirements supports PICKINGMAN’s deployment and keeps product development closely connected to operational needs.

On September 22, 2026, the AI sorting line for resource recovery from mixed municipal solid waste in Liaobu, Dongguan, jointly developed by DataBeyond and Dongshi Environment, was commissioned. PICKINGMAN began operating alongside AI optical sorting equipment.

According to published project information, the complete line can automatically sort 50 tonnes of municipal solid waste in four hours, with a waste reduction rate of approximately 35% and a resource recovery rate of approximately 25%. These figures represent the combined performance of the entire line and provide a concrete reference for evaluating the contribution of embodied robots within an industrial system. Project report

PICKINGMAN’s application needs to be understood within the context of the whole production line. Equipment layout, material conveying, and upstream and downstream processes all affect sorting performance. Aligning the robot’s tasks with the overall process turns individual machine capabilities into practical contributions to production.

For operators, the value of an intelligent sorting solution depends on effective separation of recyclables, changes to manual workflows, ease of maintenance, and the relationship between recovery revenue and operating costs.

Expanding from the Liaobu project to other sites will require adaptation to different material compositions, facility conditions and processing capacities. Continued operational experience and improvements in deployment and maintenance will support future applications. Recovering more resources and providing better technological support for frontline sorting work are the goals behind PICKINGMAN’s deployment.