<p>Robots are showing significant application potential in replacing manual workers to achieve automated belt grinding of complex components such as blisks, centrifugal impellers, and closed impellers of aero engines. However, the narrow processing channels and complex structures of the components pose a huge challenge to collision-free toolpath planning, which remains a key bottleneck greatly limiting their productivity and ground surface quality. Hence, this work presents a multi-objective planning method of collision-free machining postures for robotic belt grinding of complex components. With this method, the concepts of tool-accessible space (TAS) and conservative tool model (CTM) are first proposed to facilitate collision detection for the belt grinding process. Then, a rapid collision detection method between CTM and complex components is advanced, which enables parallel and batch computation of TAS. Based on this, a TAS-based multi-objective optimization algorithm of grinding postures with the comprehensive consideration of collision avoidance, posture smoothness, and kinematic performance of robots is further developed. The effectiveness and superiority of the proposed method are verified by posture planning experiments for robotic belt grinding of aero-engine blisk, which shows that the authors’ method can improve the computational efficiency of TAS by nearly 25 times while ensuring accuracy, and is of great significance in improving the toolpath planning efficiency for robotic belt grinding of complex components with narrow TAS.</p>