In this paper we present a new closed-loop automatic tuning methodology for decentralised proportional–integral–derivative (PID) controllers applied to two-inputs–two-outputs (TITO) non-singular processes. The main feature of the technique is the estimation of the process parameters by evaluating two closed-loop set-point step responses with two roughly tuned PID controllers already in place. Then, once a process model has been obtained, the PID controllers can be retuned by using any rule available in the literature and suitable for the application. Simulation examples, comparative results with other works and a real application on a neonatal incubator prototype are given to illustrate the methodology and to show its effectiveness