Model-Free Learning Control for Processes with Constrained Incremental Control Signal

S. SYAFIIE; F. TADEO; E. C. MARTÍNEZ

Munich, Alemania

Simposio; IEEE International Symposium on Intelligent Control; 2006

IEEE

This paper proposes a technique to design controllers for systems with constrained incremental control and input-output constraints. Model-Free Learning Control (MFLC) is a simple approach, based on Reinforcement Learning algorithms, that does not need precise detailed information on the system. MFLC is proposed for process control problems, that always present constraints. The design of a controller for a two-tank system is given as an example. Simulation results show that the proposed controller learns to control adequately the process.