You can download this project from

https://github.com/lionzheng10/LaserMeasurement

The laser measurement project is come from Udacity Nano degree course "self driving car" term2, Lesson5.

Introduction

Imagine you are in a car equipped with sensors on the outside. The car sensors can detect objects moving around: for example, the sensors might detect a bicycle.

The Kalman Filter algorithm will go through the following steps:

  • first measurement - the filter will receive initial measurements of the bicycle's position relative to the car. These measurements will come from a radar or lidar sensor.
  • initialize state and covariance matrices - the filter will initialize the bicycle's position based on the first measurement.
  • then the car will receive another sensor measurement after a time period Δt
  • predict - the algorithm will predict where the bicycle will be after time Δt. One basic way to predict the bicycle location after Δt is to assume the bicycle's velocity is constant; thus the bicycle will have moved velocity * Δt. In the extended Kalman filter lesson, we will assume the velocity is constant; in the unscented Kalman filter lesson, we will introduce a more complex motion model.
  • update - the filter compares the "predicted" location with what the sensor measurement says. The predicted location and the measured location are combined to give an updated location. The Kalman filter will put more weight on either the predicted location or the measured location depending on the uncertainty of each value.
  • then the car will receive another sensor measurement after a time period Δt. The algorithm then does another predict and update step.

Kalman filter equation description

Kalman Filter overview



2D state motion. State transition matrix, x′ = Fx + v

  • x is the mean state vector(4x1).For an extended Kalman filter, the mean state vector contains information about the object's position and velocity that you are tracking. It is called the "mean" state vector because position and velocity are represented by a gaussian distribution with mean x.
  • v is a prediction noise (4x1)
  • F is a state transition matrix (4 x 4), it's value is depend on Δt
  • P is the state covariance matrix, which contains information about the uncertainty of the object's position and velocity.
  • Q is process covariance matrix (4x4), see below
  • z is the sensor information that tells where the object is relative to the car.
  • y is difference between where we think we are with what the sensor tell us y = z - Hx'.
  • H is a transform matrix. Depend on the shape of x and the shape of y, H is fixed matrix.
  • R is the uncertainty of sensor measurement.
  • S is the total uncertainty.
  • K , often called Kalman filter gain, combines the uncertainty of where we think we are P' with the uncertainty of our sensor measurement R.

State transition Matrix F

Process Covariance Matrix Q

We need the process vovariance matrix to model the stochastic part of the state transition function.First I'm gosing to show you how the acceleration is expressed by the kinematic equations.And then I'm going to use that information to derive the process covariance matrix Q.



Say we have two consecutive observations of the same pedestrian with initial and final velocities. From the kinematic formulas we can derive the current position and speed as a function of previous state variables, including the change in the velocity or in other words, including the acceleration. You can see how this is derived below.

Looking at the deterministic part of our motion model, we assume the velocity is constant. However, in reality the pedestrian speed might change.Since the acceleration is unknown, we can add it to the noise component.And this random noise would be expressed analytically as in the last terms in the equation.

So, we have a random acceleration vector in this form, which is described by a 0 mean and the covariance matrix, Q. Delta t is computed at each Kalman filter step, and the acceleration is a random vector with 0 mean and standard deviations sigma ax and sigma ay.

This vector can be decomposed into two components. A four by two matrix G which does not contain random variables. a, which contains the random acceleration components.

Based on our noise vector, we can define now the new covariance matrix Q. The covariance matrix is defined as the expectation value of the noise vector. mu times the noise vector mu transpose.So, let's write this down. As matrix G does not contain random variables, we can put it outside expectation calculation.

This leaves us with three statistical moments. The expectation of ax times ax, which is the variance of ax, sigma ax squared. The expectation if ay time ay which is the variance of ay, sigma ay squared. And the expectation of ax times ay which is the covariance of ax and ay. Ax and ay are assumed uncorrelated noise processes. This means that the covariace sigma ax, ay in Q nu is 0. So after combining everything in one matrix, we obtain our four by four Q matrix.

So after combining everything in one matrix, we obtain our four by four Q matrix.

Program structure

  • main.cpp

    The main.cpp readin the data file, extract data to a "Measurement package". Creat a tracking instance to analyze the data.
  • kalman_filter.h

    Declere KalmanFilter class
  • kalman_filter.cpp

    Implement KalmanFilter functions, Predict() and Update()
  • tracking.h

    Declare an tracking class, it include an KalmanFilter instance.
  • measurement_package.h

    Define an class MeasurementPackage to store sensor type and measurement data.
  • tracking.cpp

    The constructor function Tracking() declare the size and initial value of Kalman filter matixes.

    The ProcessMeasurement function process a single measurement, it compute the time elapsed between the current and previous measurements. Set the process covariance matrix Q. Call kalman filter functionpredict and update. And output state vector and Covariance Matrix.
  • obj_pose-laser-radar-synthetic-input.txt

    A data file download from course web site, put it in the same folder with excutable file.
  • Eigen folder

    Library for operate matix and vector and so on. Put this folder in src folder.

Makefile structure

  • CMakeLists.txt

How to build and run this project

I am using ubuntu 16.4

  1. make sure you have cmakesudo apt-get install camke
  2. at the top level of the project repository mkdir build && cd build
  3. from /build cmake .. && make
  4. copy obj_pose-laser-radar-synthetic-input.txt to build folder
  5. from /build ./main

This is what the output looks like.

lion@HP6560b:~/carnd2/LaserMeasurement/build$ ./main
------ step0------
Kalman Filter Initialization ------ step1------
z_(lidar measure value: px,py)=
0.968521
0.40545
x_(state vector: px,py,vx,vy)=
0.96749
0.405862
4.58427
-1.83232
P_(state Covariance Matrix)=
0.0224541 0 0.204131 0
0 0.0224541 0 0.204131
0.204131 0 92.7797 0
0 0.204131 0 92.7797 ------ step2------
z_(lidar measure value: px,py)=
0.947752
0.636824
x_(state vector: px,py,vx,vy)=
0.958365
0.627631
0.110368
2.04304
P_(state Covariance Matrix)=
0.0220006 0 0.210519 0
0 0.0220006 0 0.210519
0.210519 0 4.08801 0
0 0.210519 0 4.08801

Kalman filter, Laser/Lidar measurement的更多相关文章

  1. 卡尔曼滤波—Simple Kalman Filter for 2D tracking with OpenCV

    之前有关卡尔曼滤波的例子都比较简单,只能用于简单的理解卡尔曼滤波的基本步骤.现在让我们来看看卡尔曼滤波在实际中到底能做些什么吧.这里有一个使用卡尔曼滤波在窗口内跟踪鼠标移动的例子,原作者主页:http ...

  2. 卡尔曼滤波器【Kalman Filter For Dummies】

    搬砖到此: A Quick Insight     As I mentioned earlier, it's nearly impossible to grasp the full meaning o ...

  3. [OpenCV] Samples 14: kalman filter

    Ref: http://blog.csdn.net/gdfsg/article/details/50904811 #include "opencv2/video/tracking.hpp&q ...

  4. (转) How a Kalman filter works, in pictures

    How a Kalman filter works, in pictures I have to tell you about the Kalman filter, because what it d ...

  5. [Math]理解卡尔曼滤波器 (Understanding Kalman Filter) zz

    1. 卡尔曼滤波器介绍 卡尔曼滤波器的介绍, 见 Wiki 这篇文章主要是翻译了 Understanding the Basis of the Kalman Filter Via a Simple a ...

  6. [Math]理解卡尔曼滤波器 (Understanding Kalman Filter)

    1. 卡尔曼滤波器介绍 卡尔曼滤波器的介绍, 见 Wiki 这篇文章主要是翻译了 Understanding the Basis of the Kalman Filter Via a Simple a ...

  7. [Scikit-learn] Dynamic Bayesian Network - Kalman Filter

    看上去不错的网站:http://iacs-courses.seas.harvard.edu/courses/am207/blog/lecture-18.html SciPy Cookbook:http ...

  8. 泡泡一分钟:Robust Attitude Estimation Using an Adaptive Unscented Kalman Filter

    张宁 Robust Attitude Estimation Using an Adaptive Unscented Kalman Filter 使用自适应无味卡尔曼滤波器进行姿态估计链接:https: ...

  9. 详解Kalman Filter

    中心思想 现有: 已知上一刻状态,预测下一刻状态的方法,能得到一个"预测值".(当然这个估计值是有误差的) 某种测量方法,可以测量出系统状态的"测量值".(当然 ...

随机推荐

  1. 使用Entity Framwork 保存数据时,提示不能在对象中插入重复键,违反了PRIMARY_KEY约束

    这种情况,大多发生在有外键存在的情况下,解决方法是: 把dataContext.Set<T>().Add(model)修改成dataContext.Models.Add(model);

  2. IIS Express被局域网访问

    在 文件夹 C:\Users\administrator\Documents\IISExpress\config 下面 applicationhost.config 文件里 找到相应的项目 如 < ...

  3. hdu3746 KMP-next数组的应用

    Cyclic Nacklace Time Limit: 2000/1000 MS (Java/Others)    Memory Limit: 32768/32768 K (Java/Others)T ...

  4. sql函数将1,2,3转换为表

    /****** Object: UserDefinedFunction [dbo].[splitstring_to_table] Script Date: 2017/7/11 9:35:58 **** ...

  5. java——链表、链表栈 LinkedListStack、链表队列 LinkedListQueue

    LikedList: package Date_pacage; public class LinkedList<E> { public static void main(String[] ...

  6. java——编译和运行

    Java源代码---->编译器---->Java字节码(即虚拟指令..class文件.特殊的二进制文件.二进制字节码文件)---->jvm---->解释器(jvm的一部分)-- ...

  7. java——巨简陋文本编辑器

    String :equals()方法是进行内容比较,而不是引用比较. “==”比较两个变量本身的值,即两个对象在内存中的首地址. Scanner :用Scanner实现字符串的输入有两种方法,一种是n ...

  8. mysql 存储过程(代码块)

    大纲: 创建.删除.调用. 声明变量.变量赋值 声明游标 声明异常处理器 判断 循环 使用心得 一.创建.删除.调用 创建 DELIMITER $$ #修改分隔符 )) #括号里是入参.IN代表传入的 ...

  9. if else 选择机构 _多重if选择机构_if选择结构嵌套(综合练习题——code)

    import java.util.*; class If01{ public static void main(String[ ]args){ //练习1:假如张三参加Java考试,判断如果在95分以 ...

  10. RTT之shell

    两种shell的切换:如果打开了FINSH_USING_MSH而没有打开FINSH_USING_MSH_ONLY,finsh同时支持两种c-style模式与msh模式,但是默认进入c-style模式, ...