The contact impact at the cut-in phase of the robotic belt grinding process is a critical issue that cannot be disregarded, as it is generally accompanied by contact force fluctuation and the over-grinding phenomenon, both of which affect machined surface quality and machining system stability. In this paper, a novel optimization method based on model predictive control (MPC) is presented to address the problem of cut-in contact impact, which can be directly attributed to the inappropriate transition contact state between the belt and the workpiece. Specifically, the local process parameters, including the contact force and contact velocity, as well as the trajectory characterized by the contact position prior to the cut-in stage, are dynamically optimized and planned based on the MPC-based strategy. Additionally, the instantaneous impact energy index (defined as the contact impact strength) is incorporated into the continuous task optimization objectives, which can be divided into different grinding phases, i.e., non-contact, cut-in, normal-cut, and cut-out phases. Finally, both numerical simulations and experiments on the scenario for robotic belt grinding of Ti-6Al-4V workpieces are conducted to evaluate the effectiveness of the proposed optimization methodology, indicating that the method can realize smooth contact transition and contact impact suppression.