The growing popularity of location-enabled devices has driven the demand for more efficient and reliable geographical routing algorithms for mobile ad hoc networks (MANET) and wireless sensor networks. The characteristic of this network is that nodes exchange geographical location with neighbors, and no global topology or central management is required. This paper proposes a hybrid approach for routing in MANET and wireless sensor networks without location service. Our approach combines both geographical routing and On-Demand Distance Vector Routing where location information is requested on demand. The request is used to create a backup shortest-path route to avoid dead-ends. The proposed algorithm can adopt automatically if a minimum connected dominating set (MCDS) exists, which is to limit the flooding of location and route requests. The performances metrics in this study include: packet loss, protocol overhead, end-to-end latency, and path length. Our simulation shows that our hybrid algorithm is more efficient, scalable and adaptable to rapid topology changes as compared to other popular routing algorithms, such as ad hoc on-demand vector (AODV) routing protocol and greedy perimeter stateless routing (GPSR) protocol. The hybrid algorithm combines two efficient routing algorithms where each one is a backup for the other and hence it reduces the need for flooding and overcomes the problems associated with planarization, face routing and location service. Our hybrid approach has the potential to support 3D topologies as well.