The rigid flexible coupling robotic arm combines the stable load-bearing capacity of rigid components with the flexible adaptability of flexible structures. This flexibility also leads to a complex coupling relationship between rigid components and flexible structures in the robotic arm. This coupling relationship will dynamically change with the movement of the robotic arm, causing motion disturbances and making it difficult to generate effective compensation and adjustment signals to counteract the disturbance effects and reduce control accuracy. Design an optimized control system for a rigid flexible coupled robotic arm based on the fusion of deep neural networks to address the above issues. By tracking the changes in the input signal of the robotic arm"s pose, defining its higher-order derivative, capturing the changes in the pose signal, and accurately determining the position of the robotic arm. Design a coupling module to extend the state observer to estimate the disturbance state of the robotic arm position and generate new adjustment signals. Input the signal into the motion disturbance compensator, perform feedforward operation based on the compensation parameter values, derive quasi-static processes, accurately compensate for motion disturbances, and reduce the disturbance caused by changes in the coupling relationship between rigid components and flexible structures on the motion offset of the robotic arm. Representing the motion pose of the robotic arm in three-dimensional space to solve the forward kinematics equation of the robotic arm, defining the dynamic equation of the robotic arm based on the energy dynamic distribution law of the coupled observer, and realizing the kinematic and dynamic analysis of the rigid flexible coupled robotic arm. Based on the fusion of deep neural networks, a complex mapping relationship between input and output is constructed, and multi-layer nonlinear transformation principles are defined. The input of each module of the robotic arm motion controller is used as the model input term, and the kinematic and dynamic expressions are used as the equation operation standards. The operation values of the input term are verified based on the complete neural function, and the robotic arm control conditions that meet the practical application requirements are output to complete the optimization control. The experimental results show that the designed system not only controls the end pose of the robotic arm, but also ensures the convergence performance of the motion trajectory. The response curve of the motion disturbance compensator also highlights the effectiveness of the system in controlling the joint angle of the robotic arm.