As the targets and scenarios of grasping tasks become diversified, it is difficult for traditional grasping control methods to successfully grasp targets with different materials under the influence of noises. To address the above issue, this chapter proposes a noise-tolerant grasping force optimization (NTGFO) scheme with dynamic neural network (DNN) for multi-fingered dexterous hands. The proposed NTGFO scheme ensures the optimal force to grasp the target and enables successful target manipulation in the presence of noise perturbations. This scheme consists of two main components. On the one hand, an optimal grip force model is constructed for solving the grasping force optimization (GFO) problem to keep the target from damage or sliding. On the other hand, a noise-tolerant dynamic neural network (NTDNN) is constructed to address the aforementioned model, exhibiting robustness and efficient performance in mitigating noises. Simulations and physical experiments are conducted using a Franka Emika Panda robot fitted with a three-fingered dexterous hand to assess the feasibility of the NTGFO scheme. Comparative experiments are also conducted to showcase the superiority of the NTGFO scheme.