Reliable data delivery in distributed sensor networks can be achieved by selecting error free links, prompt recovery from packet losses, and avoidance of congested gateways. Since link failures and packet losses are unavoidable in resource-constrained sensor networks, it might be necessary to tolerate a certain level of reliability without significantly affecting packets delivery performance in favor of real-time packet delivery and efficient energy consumption. This paper presents an effective cross-layer approach that improves packet delivery, maintains low packet error ratio, minimizes computation overhead, and adaptively reduces control traffic in favor of high success reception ratios of representative data packets. Based on this approach, the proposed routing scheme achieves a moderate energy consumption and high packet delivery ratio even in environments featuring high link failure rates. The effectiveness of the proposed routing scheme is experimentally investigated using a mote-based testbed as well as simulation. It is shown to be more robust and energy efficient than the network layer of TinyOS2.x. The results show that the scheme maintains higher than 95% connectivity in interference-prone channels while achieving an average of over 35% energy savings.