Robotics
4-DoF pose graph SE(2)×ℝ
ActivePublished
Key
innovation
Reducing estimation to 4 DoF (SE(2)×ℝ) under gravity alignment while keeping height as a separate variable.
Category
Robotics
Abstraction level
Pattern
Operation level
Robot control
How it works
Poses in the graph are modeled as elements of SE(2)×ℝ; edges (odometry, loop closures, inter-floor transition constraints, height priors) relate these variables, and a solver (e.g. GTSAM) minimizes the error under gravity alignment. Height remains an independent degree of freedom, allowing floors to be separated.
Problem solved
Full 6-DoF (SE(3)) optimization is over-parametrized and less stable when roll and pitch are well observed from an IMU and the key quantity is correct height (e.g. floor number).