steady-state error

Steady-state error is the gap that stubbornly remains between where you wanted a system to end up and where it actually settles, long after all the early bouncing and overshooting has died away. Picture cruise control set to 100 kilometres per hour. After the car finishes speeding up and the needle stops moving, it may hold steady at 98 — that permanent 2 kilometre-per-hour shortfall is the steady-state error. It is not about the messy first few seconds; it is the quiet, final leftover difference once everything has calmed down.

Why would a controller settle for being slightly wrong forever? Because many simple controllers only push when they can see an error, and a small error may produce only a tiny push — just enough to balance the constant drag of gravity, friction, or a slope, but not enough to fully close the gap. A controller that reacts in proportion to the error, for instance, must keep some error around in order to keep producing the output that holds the load. The classic cure is an integral term, which patiently adds up the leftover error over time and keeps nudging harder until the gap is squeezed to zero.

Steady-state error matters because it is the part of a controller's behaviour the user actually lives with day after day. A thermostat that always lands one degree cold, a robot arm that always stops a millimetre short, a level-holding tank that always sits slightly low — none of these wobble or look unstable, but each quietly fails to deliver exactly what was asked. Driving this residual to zero, or to an acceptably small value, is a core goal of good control design.