The couch in the doorway
Anyone who has tried to carry a couch through a doorway knows the feeling. The couch fits if you tilt it just so, turn it on the way in, and lift one end at the right moment — but get the order or the angle wrong and it jams. The frustrating part is that there is plenty of empty space; the trouble is purely about the couch's shape and how you rotate it as you go.
A robot faces this same puzzle, only worse. A robot arm is a chain of links and joints; a mobile robot has a body that sweeps out a footprint as it turns. Asking "can the robot move from here to there?" is never just about whether two points are far apart. It is about whether every part of the robot's body can travel the whole way without bumping into anything. Reasoning about the arm joint by joint, in the messy three-dimensional room, gets overwhelming fast.
Squashing the whole robot into one dot
Here is the trick. Instead of tracking where every part of the robot is, we ask: what is the smallest set of numbers that completely pins down the robot's pose? For a two-joint arm, that is just two numbers — the angle of the shoulder and the angle of the elbow. Once you know those two angles, the position of every link is fixed; there is nothing left to specify. Each such number is a degree of freedom, one independent way the robot can move.
Now make each of those numbers an axis on a graph. The shoulder angle runs along one axis, the elbow angle along another. Any single pose of the whole arm — links, joints, and all — becomes exactly one dot on this graph. Bend the elbow and the dot slides; swing the shoulder and it slides the other way. This abstract graph is the configuration space, or C-space, and the magic is that the robot is no longer an awkward jointed body. It is a single point.
The same recipe works for a mobile robot rolling across the floor. Its pose is fixed by three numbers: its x position, its y position, and the direction it faces. So its C-space is three-dimensional, and once again the whole robot — wheels, body, sensors — collapses to one dot drifting through that space as it drives and turns.
Where the point may not go
Collapsing the robot to a dot is only half the trick. The other half is asking what happens to the obstacles. Every pose of the robot is now a point in C-space, so we can sort those points into two piles: poses where the robot touches nothing, and poses where some part of its body is jammed into a wall, a table, or itself. This split is free space versus obstacle space. Free space is everywhere the dot is allowed to be; obstacle space is forbidden territory.
Now comes the surprise. A simple, fat pillar in the real room does not stay simple in C-space. Because the obstacle blocks the robot whenever any part of the body would overlap it, a small round pillar can carve out a large, lopsided, strangely shaped forbidden region — its exact shape depends on the robot's own body and on how that body changes as the joints turn. Deciding which side of that boundary a given pose falls on is the job of collision checking, the routine every planner calls thousands of times.
This is exactly why the doorway felt hard. In the room, the gap looks wide enough. But in the couch's C-space — its position plus its tilt angle — the doorway is a narrow, twisting corridor through free space, and you have to thread the dot along it. The robot's shape did not vanish; it moved into the geometry of the obstacle region, where a planner can reason about it cleanly.
Planning is just drawing a line
Once the world is reframed this way, the whole goal of motion planning becomes startlingly simple to state. The robot starts at one dot (its current pose) and wants to reach another dot (its goal pose). A plan is any continuous curve that connects start to goal without ever crossing into obstacle space. That is it. Find such a line and you have found a sequence of poses the robot can actually carry out.
Notice what this curve is, and what it is not. It is a path: an ordered set of poses, a route through free space. It says nothing yet about how fast to move, when to speed up, or when to slow for a corner. Adding timing and speed turns a bare path into a trajectory — that is a later step, covered when we discuss path versus trajectory. For now, the win is conceptual: a tangle of joints and obstacles has become "connect two dots with a line that stays in the white region."
Why more joints make it explode
The point trick is beautiful, but it comes with a bill, and the bill grows fast. Each degree of freedom adds another axis to C-space — another dimension the dot can move along. A two-joint arm lives in a flat two-dimensional space you can draw on paper. A typical industrial arm has six joints, so its C-space has six dimensions. A two-armed humanoid can have dozens.
Why does that matter so much? Imagine checking every pose by laying down a grid. In two dimensions, a 100-by-100 grid is ten thousand cells — easy. Add a dimension and it is a million; six dimensions is a trillion cells. Each added joint multiplies the work, so the space of poses to search balloons beyond any hope of checking it all. This blow-up is the central reason planning is hard, and it is exactly why the clever sampling and search methods in later guides exist: they find a path without ever mapping the whole space.