机器人控制算法——Bayes Filter贝叶斯滤波器
一些内容来自于:Mobile Sensing and Robotics. Link: Bayes Filter.
概率方法 Probabilistic Approaches.
主要用概率的方法,来描述,更新机器人当前的状态。需要知道全概率公式,bayes公式。其他相关的知识,不懂的补一下。
- Uncertainty in robot motion and observations.
- Use of probability theory to explicitly represent the uncertainty.
几条特性:
0 ≤ P ( A ) ≤ 1 P ( T r u e ) = 1 , P ( F a l s e ) = 0 P ( A + B ) = P ( A ) + P ( B ) − P ( A B ) P ( A ∣ B ) = P ( A B ) P ( B ) = P ( B ∣ A ) P ( A ) P ( B ) 0 \leq P(A)\leq 1 \\ P(True)=1, \quad P(False)=0\\ P(A+B)=P(A)+P(B)-P(AB)\\ P(A|B)=\frac{P(AB)}{P(B)}=\frac{P(B|A)P(A)}{P(B)} 0≤P(A)≤1P(True)=1,P(False)=0P(A+B)=P(A)+P(B)−P(AB)P(A∣B)=P(B)P(AB)=P(B)P(B∣A)P(A)
Recursive Bayes Filter.
注意一下几个notation的含义:
- belief( x t x_t xt):在 x t x_t xt时刻,机器人的位置(状态)的概率。
- x t x_t xt: state at step t t t, where the robotic are. t t t时刻,机器人的位置。
- z 1 : t z_{1:t} z1:t: observation, measurement. t t t时刻的观测值。
- u 1 : t u_{1:t} u1:t: control, action. t t t所采取的动作。
Prediction and Correction Step.
Bayes Filter can be written as two steps:
1. Prediction Step.
Motion Model.
2. Correction Step.
Observation Model.
推导笔记
下面是我做的推导的笔记(为了防止不清晰,我提供了扫描版本和原图版本),里面每一步的转换,包括全概率公式,马尔科夫假设,都没详细的去解释。如果不知道的,查一下。每一步的更新所采用的理论都进行了标注。如果有疑问,请在评论下面提出来。
- 扫描文档:
- 原图文档: