Tuning LQR Parameters using Neuro Evolution of Augmenting Topologies (NEAT) on a Double Pendulum Cart

The double pendulum cart is an example of an unstable and nonlinear system. The modeling of this system can be done using the Euler-Lagrange method with linearization using Taylor Series expansion. Linear Quadratic Regulator (LQR) is one of the optimal control techniques applied to this problem. However, the process of tuning the Q and R matrices on the LQR is often made by design. This paper proposes LQR tuning method using Neuro Evolution of Augmenting Topologies (NEAT) to ensure that the system response has good stability, cost and gain feedback that is adaptive to the state value of the system. In this paper, the best genome produced through the NEAT process scored 31.1 from the designed objective function.