Observer-controller tuning approach for double pendulum with genetic algorithm and neural network

State estimation plays a crucial role in practical control system where the use of sensors may be difficult or not economical. Observer-based controller requires the designer to be able to accurately find the gains for the observer matrix. The main objective of this work is to present a tuning technique to find gains for a Luenberger observer. The tuning technique based on genetic algorithm and neural network optimization is applied to a double pendulum stabilization problem. Sub-optimal observers have been developed for the three different unstable equilibrium positions of the pendulum. The impulse responses of the states of the system show good performance and steady-state behavior for both the methods. However, the neural network-based method is found to be superior in terms of the transient characteristics as well as gain margin.