动量蒸馏中用到了

点云配准

Jenson-Shannon divergence

B-Spline（=Basic Spline）

需求

• 需要利用样条函数拟合时相连续的数据，e.g. IMU data。
• 利用样条函数拟合散点，做表面拟合，e.g. 构建mesh或者其他光滑的表面。
• 路径规划中使用样条线对A*算法进行优化，得到更加平滑的路径。
• 多传感器融合任务中，e.g.多传感器标定时，为了内插任意时刻的状态，实现多传感器的时间对齐，使用B-spline做内插。

What is Rotation Averaging?

Problem¹

Given a set of rotation matrices \(C_i, i=\{1,2,...,n\}\), the average of them is ? Rotation Averaging is to calculate the average of a set of rotation matrices. Averaging is to get the best estimate of all rotations. (Single rotation averaging)

Or equivalently, Rotation Averaging is the problem of estimating a set of \(n\) unknown orientations \(R_1,...,R_n \in SO(d)\) from noisy measurements \(\bar{R} \in SO(d)\) of the relative rotations \(R_i^{-1}R_j\) between them. (Multi rotation averaging)

In practical the rotation averaging problem could be categoried into : Single rotation averaging、Multi rotation averaging and Conjugate rotation averaging.