Construction of Fuzzy Inference System with AI-Neuro Based Autonomous Learning Abilities and Verification of Its Effectiveness

从面向约束的角度出发，引入了模糊性的新概念，并在此基础上对模糊推理的新方法进行了形式化描述，证明了模糊推理方法能够与基于二值逻辑的人工智能中的符号处理紧密结合。然后将这种新的组合推理方法应用于不完全信息下的传感信息处理和图像理解，并应用于自主移动机器人的规划与执行。并将其应用于倒立摆的控制中，验证了其有效性。同时，利用约束逻辑程序设计语言CLP(R)为该推理系统提供了新的编程环境。利用机器学习技术，提出了一种获取模糊推理规则的学习方法，并用神经网络和遗传算法对规则进行细化和组织。最后，将具有自学习能力的模糊推理系统应用于自主移动机器人的控制与规划中，验证了该方法的有效性。

From constraint-oriented perspectives, a new concept of fuzziness is introduced, and based on this concept, new method of fuzzy inference is formalized, which is then shown to be able to be coupled tightly with symbolic procesing in AI that is based two-valued logic. This new method of combined inference is then applied to sensing information processing and image understanding under incomplete information, and also to planning and its execution of autonomous mobile robot. Moreover, the combined inference is applied to control of inverted pendulum, which elucidates its effectiveness. Also, new programming environment for this inference system is introduced by using CLP(R), a kind of constraint logic programming language. By using machine learning techniques, a learning method for acquiring fuzzy inference rules is introduced, whose results are then refined and organized by neural networks and genetic algorithms. Finally, the total system for fuzzy inference with automomous learning abilities is applied to control and planning of autonomous mobile robots, which elucidates the effectiveness of our approach.