Vehicular transportation networks are formed by interconnection of multiple networks leading to a particular class of complex networks called multi-layer networks. These networks are formed due to interaction between vehicles, vehicles and infrastructure (like a cloud), and data exchange at the infrastructure level. The interaction between these networks, and particularly the dependency of the networks formed between the infrastructure and on-road (possibly connected) vehicles at different geographical areas, affect the behaviour of all the elements in these networks. An integrated architecture of vehicular transportation networks for a multi-commodity heterogeneous traffic in an urban scenario is proposed in this paper. The heterogeneity exists due to different detection abilities of vehicles, owing to connectivity facilities and sensing capabilities, and the mode of driving (autonomous or manual) in the vehicles. Considering the multi-layer network architecture, the interconnection between the infrastructure network and the vehicle mobility network is studied. Using game theoretic tools, the impact of path prediction via infrastructure network on decision-making in vehicles and vice-versa is analyzed. The problem is formulated and modelled as a repeated Stackelberg game with incomplete information