Management of road traffic in urban settings remains a challenging problem. Perimeter control schemes proposed to alleviate congestion in large-scale urban networks usually assume noise-free measurements of the traffic state, which is problematic since measurements are corrupted by noise in reality. Moreover, for cases where estimation is employed, demand information available to the traffic state estimator is also subject to uncertainty. In this paper we develop a traffic management scheme using methods of real-time optimization based estimation and control. Firstly a nonlinear moving horizon estimation (MHE) scheme is proposed for large-scale urban road networks with route choice, with dynamics expressed using the macroscopic fundamental diagram (MFD) of urban traffic. A nonlinear model predictive control (MPC) scheme employing perimeter control actuation to minimize total time spent is then combined with the MHE to build the traffic management scheme. Case studies of congested traffic conditions in a three-region urban network showcase the potential of the MHE in providing accurate real-time traffic state information for different types of measurement configurations, leading to improved control performance under severe uncertainty regarding demands and high levels of measurement noise.