This lab works on phones, but writing code is much easier on a laptop.
Localization & SLAM · Capstone · M9-L1 · Advanced
The shuttle’s localisation has to go live on Monday. The pipeline exists in pieces: an odometry-and-lidar front end, a place recogniser and a pose-graph back end. In last week’s trial, one run ended with a bent map and nobody could say why.
You assemble the pipeline, make each piece safe against the fault it met, and then switch the defences off one at a time to learn what each is worth. The result is an engineering report you can defend.
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 is the front end of M5, M7 and M8-L3: information-form fusion with an outlier gate, graded on hidden laps with scan-match outliers and a degenerate corridor segment. Stage 2 is the loop-closure verification of M4, M5-T10 and M8-T3 with the aliased place that a fit threshold cannot catch. Stage 3 runs the whole pipeline with the Module 4 optimiser and asks for the ablation and the report of M8-T8 and M9-T2.
Problem → Experiment → Diagnose → Fix → Evaluate → Understand, on a system you assembled yourself. The grading never shows your code the true lap.
Educational simplification: SE(2), one circular lap of 24 poses, diagonal information, injected faults generated from a simple model; the occupancy map is not built and there is no timing or sensor-calibration fault. A real capstone would use recorded data, a map and the metrics of M8-T8 over many runs.
Get notified when new robotics labs are released. One email per new lab, nothing else.