SIH 2026 · PS 26053 · DRDO · Team Time2Act

Terrain perception that runs on the vehicle, with no network.

A LiDAR fires 120,000 measurements ten times a second. Keeping all of it is too slow; discarding it loses the kerb in front of you. We keep 5 cm detail where the vehicle will be in two seconds and coarsen it further out — 331× less memory, 100% of points still accounted for, 32 ms per scan on a laptop CPU.

Not a video. The Grid Lab recomputes 48,000 cells in your own browser as you move the sliders — the same algorithm, ported and tested against the Python line by line.

background: real KITTI frames, drawn live
0×
less memory than a uniform 5 cm grid
0.0%
point accuracy against human labels, 50 scans
0%
of points binned — zero data loss
0 ms
to build the map, inside a 100 ms budget at 10 Hz

The detector was never idle — the roads were smooth

Surface depressions found per scan. Identical code, identical thresholds, five drives across three sensors. The only thing that changes is the ground.

drivesensorsurfacedepressions / scan
Singapore, after rain32-beamwet tarmac0
Singapore, night32-beamarterial road6
Boston32-beamcity street36
Karlsruhe (KITTI)64-beamresidential210
Garching, off-road128-beamgravel, ruts, mud1,122

We say depression, not pothole. We checked: only 1 of 44 sat on class road — the rest are kerbs and verges. The geometry is real and correctly located; calling them all potholes would be overclaiming.

Try it — this is not a video

  1. 1Switch datasets.

    Three sensors, three countries, paved and unpaved. The same code runs all five.

  2. 2Click any cell.

    It reports its height, its size, how many points it holds and what they are.

  3. 3Move the sliders.

    In the Grid Lab the scan re-bins in your browser and the numbers change with it.

How it works

01
Raw scan

About 120,000 points per sweep from a spinning LiDAR.

02
Segmentation

SalsaNext labels every point: road, building, car, person… 20 classes.

03
Grouping

20 classes collapse into three: drivable, static obstacle, dynamic object.

OURS
04
Variable grid

Cells grow with distance. Each keeps its height — that is the "2.5D".

OURS
05
Analysis

Overhangs, surface defects, the reachable corridor, moving vs parked.

OURS
06
Advice

Slow down, steer around, or stop — the map becomes a decision.

Four rings, one origin

Resolution by distance

ringradiuscellwhy
00 – 10 m5 cmkerbs, potholes, feet
110 – 25 m10 cmvehicles, poles
225 – 50 m25 cmbuildings, lanes
350 – 100 m50 cmlayout only

Every cell size is a whole multiple of 5 cm and every ring shares one origin, so the rings tile without seams. 70.7% of a scan's points fall inside 10 m; a uniform grid spends the same memory on the empty far field.

Four drives, two sensors

Loading datasets…

The recognition model was trained on 64-beam KITTI. On the 32-beam nuScenes drives its labels are noisier; the grid itself does not change — that is the point of running it.

Honest numbers