Mei Peng, Ding GangYi, Jin QianKun, Zhang FuQuan, Jiao YangFan
Digital Performance and Simulation Technology, School of Computer Science & Technology, Beijing Institute of Technology, Beijing, China.
Fujian Provincial Key Laboratory of Information Processing and Intelligent Control, Minjiang University, Fuzhou, China.
Front Neurorobot. 2020 Dec 1;14:559366. doi: 10.3389/fnbot.2020.559366. eCollection 2020.
In this paper, we propose a creative generation process model based on the quantum modeling simulation method. This model is mainly aimed at generating the running trajectory of a dancing robot and the execution plan of the dancing action. First, we used digital twin technology to establish data mapping between the robot and the computer simulation environment to realize intelligent controllability of the robot's trajectory and the dance movements described in this paper. Second, we conducted many experiments and carried out a lot of research into information retrieval, information fidelity, and result evaluation. We constructed a multilevel three-dimensional spatial quantum knowledge map (M-3DQKG) based on the coherence and entangled states of quantum modeling and simulation. Combined with dance videos, we used regions with convolutional neural networks (R-CNNs) to extract character bones and movement features to form a movement library. We used M-3DQKG to quickly retrieve information from the knowledge base, action library, and database, and then the system generated action models through a holistically nested edge detection (HED) network. The system then rendered scenes that matched the actions through generative adversarial networks (GANs). Finally, the scene and dance movements were integrated, and the creative generation process was completed. This paper also proposes the creativity generation coefficient as a means of evaluating the results of the creative process, combined with artificial brain electroenchalographic data to assist in evaluating the degree of agreement between creativity and needs. This paper aims to realize the automation and intelligence of the creative generation process and improve the creative generation effect and usability of dance movements. Experiments show that this paper has significantly improved the efficiency of knowledge retrieval and the accuracy of knowledge acquisition, and can generate unique and practical dance moves. The robot's trajectory is novel and changeable, and can meet the needs of dance performances in different scenes. The creative generation process of dancing robots combined with deep learning and quantum technology is a required field for future development, and could provide a considerable boost to the progress of human society.
在本文中，我们提出了一种基于量子建模模拟方法的创意生成过程模型。该模型主要旨在生成舞蹈机器人的运行轨迹和舞蹈动作的执行计划。首先，我们使用数字孪生技术在机器人与计算机模拟环境之间建立数据映射，以实现机器人轨迹和本文所述舞蹈动作的智能可控性。其次，我们进行了许多实验，并对信息检索、信息保真度和结果评估进行了大量研究。我们基于量子建模与模拟的相干态和纠缠态构建了一个多层次三维空间量子知识图谱（M-3DQKG）。结合舞蹈视频，我们使用卷积神经网络区域（R-CNN）提取人物骨骼和动作特征，形成一个动作库。我们使用M-3DQKG从知识库、动作库和数据库中快速检索信息，然后系统通过全嵌套边缘检测（HED）网络生成动作模型。系统随后通过生成对抗网络（GAN）渲染与动作匹配的场景。最后，将场景与舞蹈动作进行整合，完成创意生成过程。本文还提出了创意生成系数作为评估创意过程结果的一种手段，结合人工脑电数据辅助评估创意与需求之间的契合度。本文旨在实现创意生成过程的自动化和智能化，提高舞蹈动作的创意生成效果和可用性。实验表明，本文显著提高了知识检索效率和知识获取的准确性，能够生成独特且实用的舞蹈动作。机器人的轨迹新颖多变，能够满足不同场景下舞蹈表演的需求。结合深度学习和量子技术的舞蹈机器人创意生成过程是未来发展的一个必要领域，可为人类社会的进步提供相当大的推动。
