《基于人工情感的机器人学习算法研究》由会员分享，可在线阅读，更多相关《基于人工情感的机器人学习算法研究（63页珍藏版）》请在金锄头文库上搜索。
1、哈尔滨工业大学工学硕士学位论文 I 摘 要 随着计算机和人工智能技术的发展，机器人的应用已经从传统的工业制造领域发展到了军事、民用、海洋勘探 、月球 和 火星探测等方面。这些应用领域已经不再是简单的 结构化的环境 ，而是一些极端的或者复杂多变的环境 ， 在这种情况下就对机器人的性能提出了更高的要求 。 目前的机器人的智能化水平已经无法满足上述复杂环境的 要 求，需要自主性更强，实时性更好的机器人。 传统的机器人 控制机理都是基于认知推理方式的，这种控制机理不能满足这种极端环境或者复杂多变的环境，应该从机器人控制机理上对机器人学习方式进行改进。而 关于情感智能的研究已经取得了一定的成果， 所以
2、将人工情感引入到机器人智能控制中 逐渐 成为人工智能 领域 一个新的研究方向。 本文针对传统的机器人学习算法进行研究， 将情感的因素加入到机器人学习算法中，提出了基于人工情感的机器人学习框架， 并 用于 改进 CMAC 网络学习算法、遗传算法和强化学习算法 ，提高了机器人的综合性能。 首先 ，对 CMAC 网络学习算法、遗传算法和强化学习算法 进行 分 析 ，并结合人类大脑处理信息的原理提出将情感因素引入到机器人学习中的思路，建立 了 基于人工情感的机器人学习框架，为后面分别针对 CMCA 网络学习算法、遗传算法和强化学习 算法的改进作基础。 其次， 针对自然情感的特点进行深入研究，根据情感建
3、模的知识分别建立了基于欧氏 空间 的模型和基于 认知推理的情感模型， 并通过仿真实验验证了前两种模型的有效性，证明了可以将人工情感模型应用于机器人智能控制中去。在此基础之上建立了基于模糊 推理 的情感模型，应用于后面的仿真实验中。 最后，在 Matlab 环境下，设置机器人生存实验和机器人导航实验 ，分别对基于CMAC 网络的改进算法、基于遗传算法的改进算法和基于强化学习的改进算法进行仿真实验。 仿真 实验结果验证了 基于 人工情感的机器人学习框架设计的合理性 和这种改进算法提高机器人整体性能的可行性。 关键词： 人工情感； 情感建模；机器人学习；算法优化 哈尔滨工业大学工学硕士学位论文 II
4、 Abstract With the development of computer and artificial intelligence technology, the use of robots has expanded from the traditional industrial manufacturing field to the military, civil, marine exploration, the moon and Mars detection, etc. These application fields are no longer simple structured
5、 environment, but also some extreme or complex environment. In this case, we need the robot to have higher performance, as the current level of the intelligent robot has failed to meet the needs of complex environment. The traditional robot control mechanism are based on the cognitive way, this cont
6、rol mechanism cant meet the demand of extreme environment or the complex environment, so we should improve the robot-study way from control mechanism. While the research about emotional intelligence has made some achievements, so joining the artificial emotion into the intelligence control of robot
7、has become a new research interest. In this paper, the research was carried out on the robot learning algorithm, we added the factor of the emotion into the robot learning algorithms to build the robot learning framework based on artificial emotion. Then we used it for CMAC network learning algorith
8、m, genetic algorithm and reinforcement learning algorithm respectively to improve robots comprehensive performance. Firstly, CMAC network learning algorithm, genetic algorithm and reinforcement learning algorithm were analyzed, and combined with the process of human brain processing information, the
9、 idea that putting the emotion factor into the process of robot learning was put forward .Then the robot learning framework based on artificial emotion was built, which was the foundation to improve CMAC network learning algorithm, genetic algorithm and reinforcement learning algorithm respectively.
10、 Secondly, the characteristics of the natural emotion were researched deeply. Then the emotion model based on Euclidean Space and the emotion model based on cognitive reasoning were established, and the simulation results verified the validity of two models, proving that artificial emotion motion co
11、uld be used in intelligent control of the robot. Last, on this basis the emotion model based on fuzzy reasoning was established, which was applied in the following simulation experiments. 哈尔滨工业大学工学硕士学位论文 III Finally, in the Matlab environment, the robot survival experiment and robot navigation exper
12、iment were set respectively to test three improved algorithm. Experimental results proved the designs rationality of the robot learning framework based on the artificial emotion and the feasibility of improved algorithm to improve the overall performance. Keywords: artificial emotion； emotion modeling； robots learning； algorithm optimization 哈尔滨工业大学工学硕士学位论文 IV 目 录 摘 要 . I Abstract . II 第 1 章 绪论 . 1 1.1 课题来源 . 1 1.2 课题背景及研究意义 . 1 1.3 人工情感研究现状 . 2 1.3.1 人工情感建模 . 2 1.3.2 人工情感识别 . 3 1.3.3 人工情感表达 . 3 1.3.4 人工情感机理 . 5 1.4 机器人学习研究现状 . 5 1.4.1 机器人学习方法 .
