State machines: how a robot decides what to do next

As a robot gets smarter, its code turns into a jungle of nested if-statements that nobody can follow. The cure is a state machine: a clean way to organize behavior.

The idea

At any moment the robot is in exactly one state, a mode like "following the line", "avoiding an obstacle", or "searching for a victim". Each state does one job. Transitions are the rules that move it from one state to another when something happens.

FOLLOW_LINE --(sees obstacle)--> AVOID_OBSTACLE
AVOID_OBSTACLE --(back on line)--> FOLLOW_LINE
FOLLOW_LINE --(sees silver strip)--> ENTER_ZONE

Why it helps

• One job at a time: the code for "avoid obstacle" does not have to worry about victims or intersections.
• Predictable: you can draw the whole behavior as a diagram and reason about it.
• Debuggable: print the current state and you instantly know what the robot thinks it is doing.

A Rescue Line example

A camera-based robot might have a "line" machine (following, gaps, obstacles, intersections, ramp) and a "zone" machine (find victims, pick up, deposit, find exit). A top-level mode switches between them when the robot enters or leaves the evacuation zone.

The practical tip

Give every state a name and, while debugging, show it on screen. When something goes wrong, the state tells you where to look: is it stuck in "search", flipping between two states, or entering the wrong one? A named state machine is the difference between a robot you can improve and one you can only pray over.