Optimization algorithm of an artificial neural network-based controller and simulation method for animated virtual idol characters

Abstract

To improve the movement naturalness and coordination of computer animation virtual idol characters, a virtual character interaction control method in virtual rehearsal is analyzed and proposed. Then, an artificial neural network motion controller based on motion capture data is proposed. The motion data are used as a priori information to guide the learning of the virtual character motion controller, and the whole learning process uses a progressive return function. Finally, an artificial neural network motion controller based on motion capture data uses the motion capture data as prior information and constrains the original motion capture data as part of the reward function, thereby limiting the motion pose of the virtual character. The virtual character produces a motion effect consistent with the posture of the original motion data, and the original motion data of different postures are used for experimental analysis. The experimental results show that adding spatiotemporal constraints to the return function can effectively prevent the bad motion posture of virtual characters (half cheetah and hopper) and make their motion more natural. The integration of motion data not only ensures the consistency between synthetic motion and original motion but also accelerates the learning process of network parameters.