Apart from vehicle mobility, data rate (bit rate) and multihop data transmission efficiency (including route length) have a significant impact on the performance of a routing protocol for vehicular ad hoc networks (VANETs). Existing routing protocols do not seriously address all these issues and are not evaluated for a real VANET environment. Therefore, it is difficult for these protocols to attain a high performance and to work properly under various scenarios. In this paper, we first discuss the challenges of routing in VANETs based on the data acquired from real-world experiments and then propose a routing protocol that is able to learn the best transmission parameters by interacting with the environment. The protocol takes into account multiple metrics, specifically data transmission rate, vehicle movement, and route length. We use both real-world experiments and computer simulations to evaluate the proposed protocol.