1.介绍了超声波概率模型
基于低成本移动机器人设计的超声SLAM
常见的超声波模型包括平均分布弧线模型、中线模型、高斯分布模型、经验模型
2.介绍了经验模型及其他模型
Sonar Sensor Models and Their Application to Mobile Robot Localizatio
3.介绍了超声波传感器相关结构、硬件
https://alselectro.wordpress.com/2013/03/08/arduinoultrasonic-sensor-for-distance-measurement/https://www.elprocus.com/ultrasonic-detection-basics-application/
4超声测距特性
characterization of ultrasonic sensors in an underground environment
https://baike.baidu.com/item/超声波测距/8544144?fr=aladdin （1）介绍超声测距原理
（2）超声波发射和反射和物体形状材料的关系.
（3）介绍超声波测距不精确的原因：测量物体形状、材料、多个物体、噪音
5.介绍超声波传感器的一些特性
Directed Sonar Sensing for Mobile Robot Navigation
笔记：
• Prediction: what data should the sensor produce when observing a known scene from a given position?
预测：如果已知外部环境和物体位置，传感器会产生什么数据？
• Explanation: given observed sensor data, what is the geometry of the scene that produced the data?
解释：如果已知传感器数据，是什么样的几何形状产生了这种数据？
The ideal range sensor in many people’s eyes would be characterized by a pencil-thin beam and high range accuracy independent of surface reflectance properties. We call this imaginary sensor the ray-trace scanner.
理想的距离传感器在许多人的眼睛将具有以下特点一种铅笔般细的光束，具有高度的测距精度，且不依赖于表面反射。我们称这个假想的传感器为射线追踪扫描仪。
While we do not dispute the real experimental results of researchers who have followed this approach, we feel their successes have been limited,because of the lack of physical justification for treating a real sonar as a ray-trace scanner plus a large Gaussian noise source。
而我们并不怀疑研究人员的真实实验结果，我们感到，遵循这一方法的国家的成功是有限的，由于缺乏物理上的理由来对待一个真正的声纳作为一个射线跟踪扫描仪加上一个大的高斯噪声源。
how can we use sonar？
那么我们该怎么样使用超声波传感器呢？
The approach they advocate is to characterize environment features as belonging to one of two categories:
“reflecting surfaces“ whose dimensions are larger than the wavelength, and ”diffracting objects“, whose dimensions are smaller than the wavelength.[1 R. Kuc ]
他们提倡的方法是将环境特征描述为属于以下两类之一:尺寸大于波长的“反射面“衍射物体”，其尺寸小于波长
We call those errors which occur at high angles of incidence time-delay range errors. Note that this effect is non-symmetric, and hence not Gaussian. The range measurement can be up to 19 centimeters too far, while usually not more than 1 centimeter too close (depending on the manner in which the system is calibrated. Lang et al. present remarkable error distribution plots which clearly detail this effect [85]. To put our knowledge of these errors to use for interpreting sonar data, we divide sonar returns into two classes。
我们称那些发生在高入射角时延范围内的误差为误差。注意这个效果是非对称的，因此不是高斯的。量程测量的距离可能会超过19厘米，而通常不会超过1厘米(这取决于系统校准的方式)。Lang等人给出了显著的误差分布图，清晰地描述了这种效应[85]。为了将这些错误的知识用于解释声纳数据，我们将声纳返回分为两类。
（1） Strong returns possess sufficient energy to exceed the threshold circuit promptly, and hence are very accurate in range (within 1 cen-timeter).
强回波具有足够的能量，可以迅速超过阈值电路，因此在1个centime范围内非常准确。
（2）Weak returns are characterized by time-delay range errors. These only surpass the detector threshold by the effect of a long charge-up in the threshold circuit and changing gain steps in the TVG amplifier
弱回报的特征是时间延迟范围错误。这仅仅是由于阈值电路中长时间的充电和TVG放大器中增益阶跃的变化才超过了探测器的阈值
其中：
频率20kHz~1GHz的声波称为超声波；波长=波速/频率，波速一般是340米/秒,1GHz=1000000Hz
所以的超声波的波长范围:0.00034-0.017m。
超声波系统图如下：
超声波传感器测距过程：
（1）首先由产生器产生发射波形
（2）接受器接受到回波后增强回波信号
（3）由阈值电路处理
（4）如果超过阈值，由TIMER计算时间并且换算成距离
6.介绍超声建概率栅格地图
Sv. Noykov .Occupancy grids building by sonar and mobile robot
笔记：
The angular detection range depends on the radiation directivity of a sonar transducer [6,13] as well as on the characteristics of the interface electronic blocks: working frequency, amplification of the received signal, beam width of the amplifier, trigger level
角度探测范围取决于声纳换能器的辐射方向性[6,13]以及接口电子块的特性:工作频率、接收信号的放大、放大器的波束宽度、触发电平
The ultrasonic sensor detects the closest reflecting surface inside its angular detection range, thereby indicating the presence of an empty space up to a certain distance.A single
reading r provides the information that one or more obstacles are located somewhere along the arc of circumference of radius r .
The probability model of the sonar includes:
(1) probability directional diagram P(θ) of the sonar; and
(2) probabilityestimation of the sonar measurements
超声波传感器在其角度检测范围内探测到最近的反射面，从而指示在一定距离内存在一个空的空间。一个单一的读取r提供了一个或多个障碍物位于半径r的圆弧上某处的信息。
声纳概率模型包括:
(1)声纳的定向概率图P(θ)
(2)声纳测量的概率估计
7.对基于概率栅格地图的方法进行对比
Experimental Comparison of Sonar Based Occupancy Grid
Mapping Methods
参考文献
[1] R. Kuc and M. W. Siegel. Physically based simulation model for acoustic sensor robot navigation. IEEE Transactions on Pattern Analysis and Machine Intelligence, PAMI-9(6):766–778, November
1987.
