一种模糊推理机的VLSI实现:单片专家系统

• 作 者: Masaki Togai Hiroyuki Watanabe 石桥林
• 出 处: 计算机科学
• 发表日期: 1989年04期
• 关键词: 推理机 专家系统 VLSI实现 不确定性推理 可扩充性 机器人系统 实时应用 专家知识 模糊逻辑推理 模糊集合论

我们给出了一种能处理不确定性和近似推理的推理机制的VLSI实现。为了有效和实时应用,设计是以模糊集合论中的“极大极小运算”为基础的,这种推理机制能处理不精确和不确定的知识,从而能获取专家知识并模拟其推理过程。我们采用定制CMOS技术实现了这种推理机制,并突出了简洁性、可扩充性和效率。定时模拟表明这种推理机每秒能完成约80,000次模糊逻辑推理。这类推理机的潜在应用包括指挥与控制领域中的实时决策和机器人系统的自适应命令生成等。

We present a VLSI implementation that can handle the inference mechanisms of uncertainty and approximate reasoning. For effective and real-time applications, design is based on the “maximal minima” in fuzzy set theory, which deals with inaccurate and indefinite knowledge to gain expert knowledge and simulate its inference process. We implemented this inference mechanism with custom CMOS technology and highlighted simplicity, scalability, and efficiency. Timing simulations show that this inference engine can perform about 80,000 fuzzy logic inferences per second. Potential applications for such reasoning machines include real-time decision-making in the field of command and control and adaptive command generation for robotic systems.