Optimal Control of Lane Keeping System Using Simulated Annealing and Linear Quadratic Regulator

Advanced driver-assistance systems (ADAS) are currently trending in automotive industry requiring optimal designs for achieving sensitive performance. In this paper, an optimal controller is proposed for lane keeping system (LKS) based on linear quadratic regulator (LQR). Simulated Annealing (SA) algorithm is applied to effectively provide an optimal solution for the weighting matrices that aids obtaining an optimal gain matrix for the state feedback loop. The proposed approach has been evaluated and tested as a software-in-the-loop (SIL) on a The Open Source Racing Vehicle Simulator (TORCS) which is a high-fidelity nonlinear car simulator. The developed controller has achieved better performance and proven reliability over manually tuned LQR controllers.