동적 환경에서의 양팔 로봇 상호 충돌 회피를 위한 CBF-NMPC 기반 비대칭 hybrid planning framework

초록

Dual-arm robots can efficiently perform complex tasks, such as assembly and collaboration, by simultaneously using both arms. However, when the workspaces of both arms overlap, the risk of arm collision increases. To handle these issues, this study proposes a hybrid planner that combines an optimal control-based global planner with a local planner based on Control Barrier Function and Nonlinear Model Predictive Control (CBF-NMPC). In the proposed framework, each arm adopts a different planning strategy: the main arm follows a globally planned optimal path, while the sub-arm treats the main arm as a dynamic obstacle and performs real-time trajectory tracking using the CBF-NMPC-based local planner. This asymmetric hybrid approach enables the dual-arm robot to avoid mutual collisions and perform stable, real-time tasks even in dynamic environments. Simulation results show that the proposed method operated at a control frequency of 50 Hz and avoided collisions, maintaining a minimum distance of 0.0174 m between the two arms with a safety margin of 0.01 m.

키워드