DDPG-Based Reinforcement Learning Framework for Action Path Optimization in New Media Animation Creation

Abstract

With the rapid development of the new media industry, animation creation occupies an important position in the fields of modern film and television, advertising and games. Traditional animation creation process mostly relies on manual operation, which is inefficient and flexible, especially in the design and optimization of action path. In order to solve this problem, this study proposes an intelligent control and optimization scheme using the Deep Deterministic Policy Gradient (DDPG) algorithm to optimize the action path in new media animation creation. The method constructs a reward function considering path fluency, precision, creation cost, and user preference, and applies a continuous control strategy within a reinforcement learning framework. We collected 1000 animation scene data samples and compared the proposed method against traditional optimization techniques including Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Simulated Annealing (SA). Experimental results show that our method reduces the action path error (MSE) from 0.082 to 0.045 (a 45.1% improvement), increases fluency from 0.87 to 0.97 (a 11.5% increase), and reduces optimization time by 55% compared with GA. The DDPG-based approach also demonstrates faster convergence and better stability. These findings confirm the effectiveness and efficiency of reinforcement learning in enhancing intelligent animation production. The research results of this paper provide a new idea and method for new media animation creation, which can greatly improve the automation degree and quality of animation production, and provide theoretical support and practical guidance for the intelligent animation creation in the future.