【论文阅读记录】Real-Time Correlative Scan Matching

这篇文章是谷歌的Cartograph中实现real_time_correlative_scan_matcher的论文

 

Real-Time Correlative Scan Matching
Edwin B. Olson
University of Michigan
Department of Electrical Engineering and Computer Science
Ann Arbor, MI 48109
Email: ebolson@umich.edu
http://april.eecs.umich.edu

提出了一种概率驱动的扫描匹配方法。

Fig. 2. Graphical model for probabilistic scan matching. xi􀀀1 represents the
previous position of the robot, u the motion of the robot, xi the new robot
position, m the world model, and z the laser scan observation.

 

we can the compute the
conditional probability that the sensor observes a nearby point
p given that mi was the cause of that observation. We repeat
this process for each point in the map, recording the maximum
probability for each point in the lookup table

使用传感器的已知信息，大概估计了机器人会往哪走，然后通过

粗暴建立了高斯分布的概率的lookup表，计算时候用GPU加速。