Abstract: As the foundation for intelligent transportation system (ITS), vehicular ad hoc network (VANET) has been receiving more and more attentions in recent years. Due to the high mobility and uneven spatial distribution of vehicles in vehicular networks, network topology changes frequently, which makes the routing links easy to break. Hence, how to design the reliable and efficient routing protocols for VANET is becoming a challenging task. To deal with this problem, this paper proposes an ant colony optimization (ACO) based delay-aware routing protocol (ACDAP) for urban VANETs. Firstly, according to urban road topology, the data forwarding process is divided into straight road mode and intersection mode. In the straight road mode, packets are transmitted along the road by means of greedy strategy. While in the intersection mode, some special nodes, named as bridge nodes, are selected at the junctions to connect the partition networks over road segments. These bridge nodes are responsible for launching small control packets termed as ants. Meanwhile, these bridge nodes calculate the weights of segments by using a delay-aware method that is designed according to the pheromone updating mechanism and the exploration of the optimal path in ant colony algorithms. These weights can well reflect the connectivity of segment networks. Based on the calculated weight for each road segment, a reactive routing scheme is proposed to obtain the optimal route with high network connectivity. Finally, the real urban vehicular scenario is simulated via the NS2 simulation platform, and the comparison between the ACDAP with several routing protocols including GPSR, GSR and VACO is made, from which it is shown that the proposed ACDAP protocol in this work outperforms the conventional protocols in terms of in both data packet delivery rate (PDR) and end-to-end delay (E2ED).