Active Interaction Framework on Door-Pushing Tasks for Ducted Fan Aerial Platform with Multi-dimensional Perception

Door-pushing tasks are increasingly critical for aerial robotics to accomplish inspection and rescue missions, yet presents significant challenges in practical applications regarding integrated perception-to-control. We introduce a novel multidimensional feature recognition method for traversability assessment and establish the holistic active interaction framework from perception to control. The perception algorithm, the first known complete multi-dimensional method to our knowledge, is open-sourced at https://github.com/Jadeiin/AeroInteract. Simulations and real-world experiments carried out validate the proposed solution, achieving successful autonomous door traversal while demonstrating generalization capability of the perception algorithm across diverse door materials.