This lab works on phones, but writing code is much easier on a laptop.
Localization & SLAM · Visual and Visual-Inertial SLAM · M6-L3 · Advanced
The robot lost tracking three times on the same patrol. The vendor says “the visual SLAM is unreliable”. The logs say something more useful: each time the cause was different.
You read the logs like an engineer: classify each loss, then decide when a recovery should be believed.
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 applies the failure catalogue of M6-T9: every cause leaves a fingerprint in a different signal, and the grader checks the start and the cause of each episode on hidden logs with different combinations of causes. Stage 2 is the recovery side: how inlier count and persistence decide whether a relocalisation is real, and the delay each check costs.
Each cause has a different remedy (shorter exposure, another sensor, locked exposure, a design change, masking). Without the diagnosis you tune the wrong thing.
Educational simplification: the logs are synthetic per-frame summary signals; each episode has a single cause, a fixed length and clean separation of signals; relocalisation proposals are described by two numbers. Real logs mix causes and need thresholds tuned on data.
Get notified when new robotics labs are released. One email per new lab, nothing else.