The spatio-temporal routing in episodically connected vehicular networks es a communication pattern where the destination nodes is a set of vehicles that transit the destination region during the specified destination region. The spatio-temporal routing is well suited to support Smart City and Intelligent Transportation Systems applications (i.e. road safety, crowdsensing, entertainment). In this paper, we present Oportunistic Greedy Routing over Street-layout Graph (OGRoSG), a novel spatio-temporal routing protocol for episodically connected vehicular networks. Unlike previous works in the literature, OGRoSG takes advantage of information available in modern navigation systems, such as the streets map and the geographic position. The protocol performs greedy geographic routing, the notion of being closer to the destination area is based on the progress towards the destination region on the streets-map graph and the direction of the vehicles. In order to achieve this, the vehicles process the destination region and the street graph to generate a shortest-paths tree with an extended destination region as root. This extended destination region guarantees that it covers the entire destination region and the paths that interconnect it. We evaluate the performance of the OGRoSG protocol using simulations in NS-3, using mobility traces generated in a map based in the city of Murcia, Spain. Our results show that OGRoSG outperforms a set of Spatio-Temporal variants of the Epidemic, Spray & Wait and Binary Spray & Wait routing protocols in terms of delivery ratio and overhead induced in the network.