Signal Control for Urban Traffic Networks with Unknown System Parameters

Among the several signal control strategies that have been proposed in the literature, a key assumption is that system parameters including network service rates and demands are known. However, it is envisaged that in the next generation of transportation networks with mixed autonomy, system parameters such as service rates may vary as autonomous vehicle penetration rate changes. Aligned with this, we propose a signal control strategy which, unlike previous approaches, can handle both unknown mean network demands and service rates. To this end, we use stochastic gradient projection to develop a cyclic iterative control, where at every cycle, the timing plan of the signals is updated. At each iteration, the update rule is based on the measured changes in the network queue lengths. If the network mean arrival and service rates are assumed to be constant, the proposed iterative signal control is guaranteed to converge to an optimal solution. We describe the intuition behind our algorithm, and further demonstrate through simulation studies that our iterative control scheme can successfully stabilize the system.