Motion Planning and Nonlinear Model Predictive Control for Dual-Arm Collaborative Manipulation

Precise dual-arm manipulation is essential for robotic applications. Despite the dexterity and stability in bimanual tasks, dual-arm manipulation remains challenging due to complex kinematics, strong nonlinearities, and multiple nonconvex constraints. This paper addresses these challenges from both planning and control perspectives. First, a motion planning method based on a sample-based method is proposed to incorporate task constraints and avoid collisions between arms and the environment. Additionally, it enables determining the initial and goal configurations directly from task-space constraints, avoiding the need for explicit joint-space targets. Second, a nonlinear model predictive control (NMPC) framework is proposed to coordinate dual-arm motion by tracking the full position and principal orientation of each end-effector, while effectively handling kinematic coupling and joint constraints in real time. Finally, the proposed methods are validated via simulation, demonstrating their effectiveness and compliance with precision requirements.