Path planningSep 2026 · GirlsWhoML × PhysicsX hackathon · Team

ROVERA

From any point on Mars to the nearest water, priced in sols.

A route-planning engine that searches space and time for the fastest safe traverse to water under terrain, daylight and dust constraints, paired with a mission-control web app: a GLSL-shader 3D Mars, a FastAPI backend and AI briefings streamed live.

Read the planner code (opens in a new tab)Skills it shows

rovera · mission control

Route planner: Jezero Crater to the Utopia Planitia glacier, priced at 202 sols.

RoleBuilt the planning engine and the mission-control app

TeamHackathon team, Vehicles & Mobility track

WhenSeptember 2026, one evening

Built withPython, React, Three.js / GLSL, FastAPI

Skills it shows

Search algorithms

A* over (row, col, sol). Each sol has a drive budget, so the best route can wait out a dust storm instead of driving into it.

Modelling and simulation

MOLA-style elevation and AI4Mars-style terrain classes make the cost map; MARCI-style dust opacity drives a solar power model.

3D graphics

A custom GLSL shader planet in react-three-fiber: noise terrain, a day/night terminator and an atmosphere rim. Click it to set a start point.

Full stack and LLMs

FastAPI and MongoDB behind the app; OpenAI briefings in three personas, streamed token by token over server-sent events.

Problem, build, result

1Problem

Nearest water is not soonest water. A solar rover’s range depends on dust, daylight and the ground under it, so ranking by distance picks the wrong deposit.

2What I built

A planner that answers in sols: slopes and rocks as hard limits, a dust-driven daily drive budget, and A* over position and time. Then a mission-control app around it.

3Result

In the demo run the planner reached water in 9.3 sols, holding still for 4 sols through a dust storm, and ranked every reachable deposit by time.

9.3 solsto water in the demo run

4 solsheld through a dust storm

2 hoursfrom scope lock to push

3AI briefing personas