Designing a QoS-aware, yet energy-saving routing protocol for WSNs is a notoriously hard problem. However, the outstanding interest for this technology, and the growing number of envisioned applications, motivate the need to introduce the notion of Quality of Service (QoS) in these networks. This paper introduces EDEAR (Energy and Delay Efficient Adaptive Routing), an adaptive routing algorithm based on route exploration and reinforcement learning. We evaluate EDEAR with simulations, under various network mobility conditions. Our results show that EDEAR outperforms any other routing protocol, delivering packets with the shortest delay, while reducing energy consumption. As a result, EDEAR's features allow to increase the network lifetime by 9-18%