← Saakshi Gadia ยท Portfolio

Lunar Path Determination

Hazard-aware rover navigation. A fresh lunar terrain is generated in your browser; A* plans the shortest route and a hazard-aware safe route, and both are compared.

Shortest pathSafe pathHazardStart / goal
ShortestSafe

Browser version: hazard scores here come from slope and roughness rules so it runs instantly. The Python project trains a Random Forest on six terrain features (F1 0.88, ROC-AUC 0.98 on unseen maps). Across 20 test maps the safe planner cut hazards crossed from 29.1 to 0.3 and max slope from 35.4° to 11.9° for 14.9% extra distance.

Source code on GitHub

Result from the Python pipeline (trained model):

Shortest vs safe path from the trained model