机器人学习--Robotics: Estimation and Learning（宾夕法尼亚大学COURSERA课程）

COURSERA网课网址：https://www.coursera.org/learn/robotics-learning#syllabus

总共四周课程，四个模块内容。

博客园网友截图及学习笔记：https://www.cnblogs.com/ecoflex/p/9868474.html

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 https://blog.csdn.net/jinshengtao/article/details/81294285

 

 

 https://www.getit01.com/p2018012421648507/

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

  

converge to a local optima

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

The robot will not directly measure X unfortunately, but

 

the robot may observe portions of x through it's sensors.

 

This portion is labeled z, where the relationship between the state and

 

measurement is given by the mixing matrix, c.

  

 

 

Creditly both X and Z contain noise even in this model.

 

State X is noisy because the linear model does not capture all

 

physical interactions.

 

Observation Z are noisy because sensors contain noise in their measurements.

 

  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

In the next section, we will show how these two probability distributions

 

can be fused to provide a better estimate of the true state.

 

 

 

 

 

 

 

 

 

 

From the dynamical system, the probability of the state given only the previous

 

state can be represented with the prior information alpha.

 

Representing the information from our measurement model,

 

beta provides observational evidence.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

building a semantic map requires

 

advanced object recognition techniques which go beyond our scope.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

But we are most interested in the range sensor it has on the top.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 https://www.cnblogs.com/flash3d/archive/2012/01/30/2332121.html

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 https://www.quora.com/Why-do-self-driving-cars-use-LIDAR-instead-of-depth-cameras-like-Kinect

 

 https://www.zhihu.com/question/28749424

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

k-d tree算法

https://www.cnblogs.com/eyeszjwang/articles/2429382.html

 

 

 

 

 

 

 

 

 

 

A

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Here, theta encompasses 270 degrees, not a full circle.

  

The laser scanner can only see 10 to 30 meters away.

  

In this range restriction, means that distance measurements

 

showing here is black dots, can only be found within the area in green.

 

 

 

 

 

The complimentary stages of mapping and localization when performed together

 

are known as SLAM, simultaneous localization and

 

mapping, which is a major research topic in robotics.

 

 

 

 

 

 

 

  

we have some tricks to make the search easier.

 

We can constrain the search to a limited number of poses based on odometry information.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

http://www.cnblogs.com/maybe2030/p/5043356.html

 

 https://blog.csdn.net/heyijia0327/article/details/40899819

 

 https://www.cnblogs.com/21207-iHome/p/5237701.html

 

 

 

 

 

 

 

 

 

    

 

    

 

 

so, we need to find a rotation and translation that move the measured points to match the model points.

 

         

 

The strategy of the ICP algorithm takes an optimistic assumption that the point sets are close enough.

 

 

 

 

 

 

 

       

 

 

 

 

知乎上网友的学习笔记：https://zhuanlan.zhihu.com/p/21648507?refer=robotics-learning

建图部分的网友解析：https://www.cnblogs.com/ecoflex/p/9868474.html