Implement EKF
Description
The extended kalman filter is a standard algorithm for state estimation. This can then be used for tracking, prediction, or even SLAM.
Tasks
High level overview of packages to introduce:
-
Motion model -
catr
model -- constant acceleration and turning rate -
Constant acceleration model -
Motion model interface to be used by kalman filters
-
-
Math package - Current choice is eigen
-
autoware_blas
- A future issue will replace this custom blas implementation with an open source alternative, hopefully from eigen
-
Message abstraction layer- This task became outside of the scope for this issue, since the EKF is implemented as a library; the node that leverages the EKF will need a message abstraction layer
-
Integrate with autoware_cmake
After the initial implementation (task list above), the team decided that it will be more valuable to the community to remove autoware_blas
(for more info see #20 (closed)) and switch to Eigen now.
-
Replace autoware_blas
with Eigen3 -
Remvoe autoware_blas
from the source code
Edited by Lyle Johnson