Autonomous Cave Rescue: Frontier Exploration & Multi-Robot Path Planning (2D)

Cave Generation

Obstacle density 42%

Rescue Fleet

Number of robots 2

LIDAR sensor range 8.0 cells

Robot speed 3.0 cells/s

Live Readout

Cave mapped

Survivors found

0 / 0

Elapsed time

0.0 s

Robots active

0 / 0

How it works

The cave is generated with a cellular automaton (random wall fill, then four smoothing passes using an 8-neighbour majority rule), then the largest connected floor region is kept via flood fill so it is always fully solvable. Robots know nothing about the layout at start — everything is fog of war.

Each robot casts 48 simulated LIDAR rays every frame, marking cells free or wall in a shared occupancy grid and revealing any survivor it gets an unobstructed line of sight to within sensor range. When a robot has nowhere queued to go, it runs a breadth-first search outward from its own cell across known free cells only, and returns the nearest cell that borders unexplored territory — a frontier — skipping any frontier another robot has already claimed, so the fleet spreads out instead of piling onto the same target. The BFS also hands back the shortest known-safe path to that frontier for free.

Fog — unexplored cells; nothing is assumed about them.
Frontier BFS — re-run every time a robot finishes its current path, so exploration always chases the closest unclaimed unknown boundary.
Mission end — a robot stops when no unclaimed frontier is reachable from what it currently knows; the run ends when every robot stops or every survivor is found.

Exploring…

Unexplored (fog) Mapped floor Mapped wall Trapped survivor (found)