Within the past few years several promising centralized control optimization strategies with adaptive capabilities and event-based traffic regulation strategies have been proposed based on dynamic programming principles, presenting increased efficiency, maximized serving capability and quality of service in this area (improved mean network speed, traffic demand service and improved travel times). Moreover modern optimization approaches have also been proposed in literature presenting very promising results. Unluckily most of these approaches depend heavily on the model availability and precision and formulation (linear, nonlinear) which in many practical applications might not be available. To bypass such problem, oversimplified system models are adopted to ensure applicability in many practical applications. As a result the real-life plant is optimized based on imprecise models, which usually puts efficiency at stake.

To this end, novel centralized and decentralized adaptive optimization procedures – namely Cognitive Adaptive Optimization (CAO) and Local for Global Cognitive Adaptive Optimization (L4GCAO) respectively – have been developed aiming towards enabling cognitive, adaptive and optimized regulation of the adopted control strategy, in very large scale complex traffic control systems with unknown dynamics. Both CAO and L4GCAO accomplishes to perform efficiently even in cases when the network model is poor or unknown. CAO approach has already been tested and evaluated in similar, simulation and real-life, applications presenting high operational efficiency.