基于认知体系的自适应行为建模及应用研究
文献类型：学位论文
| 作者 | 王雪 |
| 学位类别 | 硕士 |
| 答辩日期 | 2017-05-24 |
| 授予单位 | 中国科学院沈阳自动化研究所 |
| 授予地点 | 沈阳 |
| 导师 | 库涛 |
| 关键词 | 自适应行为 认知体系 SOAR 移动机器人 |
| 其他题名 | Adaptive Behavior model of SOAR-based and Application Research |
| 学位专业 | 控制理论与控制工程 |
| 中文摘要 | 自适应行为的自主管理过程和自主目标实现，是目前人工智能领域的一个重要研究方向，是智能系统发展的必然趋势。虽然目前很多智能系统都具有一定的自适应能力，都能完成某种具体的任务问题，但是大多数智能系统缺少解决通用问题的能力，自适应能力较差，与人类智能还有很大差距，采用智能算法实现的自适应行为虽然提高了系统的智能性，但依然存在很多的局限性。鉴于上述存在的问题，本文将充分借鉴基于SOAR（State Operator And Result，SOAR）的认知体系研究方法，将人类的认知过程融入到自适应行为选择中，设计并实现基于认知体系的自适应行为模型，并通过问题空间搜索方法实现在自适应行为模型基础上的任务问题求解。首先需要将具体的任务问题抽象到特定的问题空间中，然后根据任务要求在特定的问题空间中设计合适的操作符，利用操作符实现状态的转移，达到预期的目标状态，实现任务的解决。为了验证自适应行为模型和问题空间自适应搜索求解方法的有效性，本文面向移动机器人在未知空间中的移动过程进行模型及方法的实验仿真验证，通过对机器人移动空间中障碍物的随机分布假设，实现机器人移动过程中的自适应路径搜索，无碰撞地达到指定的目标位置。同时，为了提高问题的解决效率，将强化学习方法融入到机器人移动行为过程中，提出了基于强化学习的路径选择算法，设计了基于执行反馈的奖赏函数，将执行过程中的反馈信息以奖赏的形式融入到奖赏函数中。仿真实验结果表明，基于强化学习和过程反馈的移动机器人自适应行为路径选择过程更加高效，实现了机器人利用环境奖赏的路径选择优化，提高了问题解决效率。 |
| 英文摘要 | Adaptive behavior of the self-management process and the realization of independent goals, is an important research direction of artificial intelligence, is the inevitable trend of the development of intelligent systems. Although many intelligent systems have a certain degree of adaptive ability, can complete a specific task problems, but most of the intelligent system lacks the ability to solve common problems, adaptive ability is poor, and human intelligence is still a big gap.The adaptive behavior based on intelligent algorithm improves the intelligence of the system, but there are still many limitations. In view of the above problems, this paper will use the cognitive system research method based on SOAR (State Operator And Result, SOAR) to integrate the cognitive process of human into the adaptive behavior choice, design and implement the adaptive behavior model based on cognitive system and use the method of the problem space search method to realize the task problem solving based on the adaptive behavior model.First, we need to abstract the specific task problem into the specific problem space, and then design the appropriate operator in the specific problem space according to the task requirement, use the operator to realize the state transfer, achieve the expected target state and realize the task. In order to verify the validity of the adaptive behavior model and the problem space adaptive search method, this paper presents a simulation of the model and method of the mobile robot in the unknown space. Through the stochastic distribution hypothesis of the obstacle in the robot's moving space, to achieve the robot mobile process in the adaptive path search, to achieve a collision-free to reach the specified target location. At the same time, in order to improve the efficiency of the problem, we will assimilate the reinforcement learning into the behavior of the mobile robot, and propose a route selection algorithm based on the reinforcement learning. We design the reward function based on the feedback and assimilate the feedback information into the reward function in the form of reward. The simulation results show that the adaptive behavior path selection process of mobile robots based on reinforcement learning and process feedback is more efficient, and the path selection optimization of robot use environment reward is realized and the efficiency of problem solving is improved. |
| 语种 | 中文 |
| 产权排序 | 1 |
| 源URL | [http://ir.sia.cn/handle/173321/20524]  |
| 专题 | 沈阳自动化研究所_数字工厂研究室 |
| 推荐引用方式 GB/T 7714 | 王雪. 基于认知体系的自适应行为建模及应用研究[D]. 沈阳. 中国科学院沈阳自动化研究所. 2017. |
入库方式： OAI收割
来源：沈阳自动化研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。