Adaptive optimization with satisfactory transient performance for large-scale traffic control systems

The majority of practical Large-Scale Traffic Control Systems (LSTCSs) requires the optimisation (finetuning) of their design parameters. A tremendous amount of human effort and time is spent for optimisation of the overall LSTCS, which is usually performed by experienced personnel in the lack of an automated well established systematic approach. Methodologies based on stochastic approximation principles have been proposed, such as the Random Directions Kiefer-Wolfowitz (RDKW) and the Simultaneous Perturbation Stochastic Approximation (SPSA). Both algorithms possess the serious disadvantage of not guaranteeing satisfactory transient behaviour, due to their requirement for using random or random-like perturbations of the parameter vector. When these methods are applied to closedloop controller optimisation applications of LSTCS, the use of random or random-like perturbations may lead to severe poor performance of the overall system, stability problems or even safety problems. A new algorithm called Adaptive Fine-Tuning (AFT) is proposed for alleviating this problem. The approach of AFT is based on a recently proposed Adaptive Optimisation (AO) methodology that is aiming at replacing the manually-based optimisation by a fully automated procedure and is proven using rigorous mathematical arguments to provide with safe and reliable, efficient and rapid optimisation of general LSTCSs. Application of the proposed scheme to the adaptive optimisation of a large-scale, complex control system demonstrates the efficiency of the proposed scheme.