This lab works on phones, but writing code is much easier on a laptop.
Localization & SLAM · Visual and Visual-Inertial SLAM · M6-L4 · Advanced
The visual-inertial system is accurate on the bench and drifts on the robot. Its estimated gyro bias flips sign with every turn, a pattern that no real bias has.
You measure the time offset between the camera and the IMU, apply it where the gyro is integrated, and find why the bench never showed the problem.
You have passed every stage. Now connect it back to the theory and move on:
Plain Python with math and, for matrices, slamkit (matmul, transpose, solve, inverse, identity). The data arrive as the dictionary D. There is no NumPy, random, json or copy in the browser runtime. Your code runs locally and never leaves this page.
Stage 1 estimates the offset by sliding two angular-speed signals against each other, graded against the hidden true offset. Stage 2 uses the offset where it matters, in the integration window between two exposures, graded against the true rotation. Stage 3 sizes the damage: pixels of disagreement as a function of offset and rotation intensity (M6-T8).
Estimators explain what they are given. An unmodelled delay becomes a fictitious bias, and the system seems fine until the robot rotates quickly.
Educational simplification: planar yaw only, a constant offset, smooth sum-of-sinusoids motion, white gyro noise and magnitude signals for the offset search. Real systems estimate the offset in 3D alongside biases and extrinsics, and the offset can drift.
Get notified when new robotics labs are released. One email per new lab, nothing else.