AIROBO, CAS Institute of Mechanics Sign Five-Year Deal on Robot Operations Research

The agreement, which the Chinese tech news outlet said is not centered on any single robot model, targets the shared technical challenges that arise when robots from multiple vendors have to work in the same physical environment. Under the pact, Aibono and the CAS institute will jointly develop capabilities covering multi-brand robot access, map and spatial modeling, task orchestration and collaborative scheduling, elevator and door access coordination, status monitoring, anomaly alerts, data-loop feedback and technology transfer.

The announcement describes a problem that becomes visible as robot fleets grow. A single delivery robot can usually be managed with the equipment maker's own software, Leiphone said. When a site has dozens or hundreds of machines from different brands, complications multiply: the robots may use different map systems, coordinate frames, task interfaces and data formats, and they also need to interact with elevators, access gates, corridors and the building's systems.

The participation of the CAS Institute of Mechanics brings engineering mechanics into what could otherwise be classified as software development. The partners plan to study complex-environment movement with contact mechanics, multi-physics coupling modeling, robot dynamics and control, dynamic obstacle prediction, multi-robot cooperation, human-robot safety in shared spaces, embodied-intelligence scene learning and scientific machine learning. The institute's expertise is meant to help the operations platform move beyond device management by adding spatial understanding, movement prediction and decision-making capabilities.

The cooperation also includes work on spatial physical modeling, simulation of real scenes, digital twins, wheeled-leg composite motion control and high-density group coordination. According to Leiphone's account, conventional maps mainly answer where a robot may go. The planned research would try to address more demanding operational questions, such as when a corridor is most crowded, how to reassign tasks after a passage becomes unavailable, and how robots should be sequenced when elevator capacity is capped. If building spaces can be digitized and continuously connected to real-time operating data, the platform may eventually simulate strategies in a digital environment and send optimized instructions to robots in the field, shifting from handling problems after they occur to predicting them in advance.

The two sides also plan to set up a robot-operations project laboratory. In the report's description, the laboratory would provide a vehicle for a closed loop running from actual operation and data collection to problem discovery, simulation, algorithm optimization and verification back on site. Because real-world environments produce long-tail events that are difficult to predefine in a laboratory, the platform's value is not merely in collecting more data but in turning field incidents into repeatable and verifiable research samples.

According to the report, AIROBO's platform can be understood as having three layers: connecting robots from different brands, coordinating their movement and access to building infrastructure, and, at the highest level, using spatial models, real-time data and algorithms to dynamically schedule tasks, allocate resources and make risk-based adjustments. The broader point made by the report is that the robot industry is moving from single-machine intelligence to group and spatial intelligence; the body determines what an individual robot can do, while the operations platform determines whether a fleet can continue to function in the physical world.