Abstract
The AI-Fleet project is developing an AI-based fleet manager for the planning and control of multi-robot systems on dynamic construction sites. The fleet manager enables effective collaboration between heterogeneous robots – including mobile platforms, autonomous construction machines, and cranes – to safely, efficiently, and adaptively execute complex construction processes. Key innovations include capability-based task allocation, a site representation supported by semantic knowledge graphs and ontologies, as well as voxel-based spatio-temporal models for dynamic path planning. The fleet manager integrates AI planning, real-time monitoring, and standardised interfaces to ensure task distribution, disturbance management, and continuous optimisation. The solution addresses central objectives such as resource efficiency, productivity enhancement, and resilience in the construction industry and will be validated in a virtual as well as a real testbed with real robots. The results will be made available as open science software and as a demonstrator for integration into CARE.
