基于模糊神经网络的多移动机器人自学习协调系统

研究多移动机器人的运动规划问题．针对机器人模型未知或不精确以及环境的动态变化，提出一种自学习模糊控制器（ＦＬＣ）来进行准确的速度跟踪．首先通过神经网络的学习训练构造ＦＬＣ，再由再励学习算法来在线调节ＦＬＣ的输出，以校正机器人运动状态，实现安全协调避撞

A base-emitter self-aligned multi-finger Si1-xGex/Si power heterojunction bipolar transistor

With a crystal orientation dependent on the etch rate of Si in KOH-based solution, a base-emitter self-aligned large-area multi-linger configuration power SiGe heterojunction bipolar transistor (HBT) device (with an emitter area of about 880 mu m(2)) is fabricated with 2 mu m double-mesa technology. The maximum dc current gain is 226.1. The collector-emitter junction breakdown voltage BVCEO is 10 V and the collector-base junction breakdown voltage BVCBO is 16 V with collector doping concentration of 1 x 10(17) cm(-3) and thickness of 400 nm. The device exhibited a maximum oscillation frequency f(max) of 35.5 GHz and a cut-off frequency f(T) of 24.9 GHz at a dc bias point of I-C = 70 mA and the voltage between collector and emitter is V-CE = 3 V. Load pull measurements in class-A operation of the SiGe HBT are performed at 1.9 GHz with input power ranging from 0 dBm to 21 dBm. A maximum output power of 29.9 dBm (about 977 mW) is obtained at an input power of 18.5 dBm with a gain of 11.47 dB. Compared to a non-self-aligned SiGe HBT with the same heterostructure and process, f(max) and f(T) are improved by about 83.9% and 38.3%, respectively.