Cooperative manipulation control with task-prioritized real-time optimization for free-floating dual-arm space robots

Dual-arm space robots, offering superior dexterity and enhanced target manipulation abilities compared to their single-arm counterparts, represent a critical technology for advanced on-orbit operations including the construction and maintenance of large-scale structures. However, their application is hindered by two key challenges: (1) strong dynamic coupling between the base and robotic arms, which is often compounded by significant variations in inertial properties; and (2) complex physical constraints, including limits on internal wrenches at the grasping points and geometrical constraints for self-collision avoidance. To address these challenges, this paper proposes a robust, real-time, task-prioritized control framework based on Hierarchical Quadratic Programming. The framework integrates an efficient neural network model to provide differentiable distance predictions, facilitating the linearization of collision constraints within a two-level structure that strictly prioritizes safety. Additionally, an online error correction mechanism is developed to counteract error accumulation and disturbances. Numerical simulations substantiate the framework’s superior computational efficiency and tracking precision, demonstrating a 1 kHz real-time control frequency with median errors of approximately 2×10⁻³ m/rad. Furthermore, the framework exhibits exceptional robustness against diverse trajectories and large variations in system inertial properties.