terrain traversability

Terrain traversability is a robot's judgement about whether a patch of ground ahead can actually be crossed — safely, without getting stuck, tipping over, or wrecking itself. It is the machine version of the quick glance a hiker gives the trail before each step: is that slope too steep, is that mud too soft, is that gap between the rocks too wide for my legs? The answer is not simply yes or no but a matter of degree, and it depends heavily on what kind of robot is asking. A staircase is an easy obstacle for a legged robot, a hard wall for a wheeled one, and irrelevant to a drone, so traversability is always judged for a particular body, not in the abstract.

To make that judgement, the robot reads the shape and feel of the ground from its sensors — usually a depth camera or a laser scanner that returns a cloud of 3D points — and then scores small cells of the surface on a few practical qualities. Slope is how tilted the ground is: gentle is fine, steep risks rolling backward or sideways. Roughness is how bumpy or uneven it is: smooth pavement is easy, a field of loose cobbles jolts and snags the wheels. Obstacles are the things sticking up or cut down — boulders, curbs, ditches, stairs — that must be climbed, stepped over, or steered around. Surface properties are what the ground is made of: firm rock grips well, while sand, snow, ice, or wet leaves can let the robot slip, sink, or bog down even where the shape looks perfectly flat.

The result is usually drawn as a traversability map, a kind of colored grid laid over the world where green means go, red means avoid, and shades in between mean proceed with caution. The planner that decides the robot's route then treats this map like a hiking map with no-go zones marked, steering around the red and threading through the green. Getting this right is the difference between a delivery rover that confidently rolls down the sidewalk and one that face-plants off a curb or gets hopelessly mired in a flowerbed.