Wireless Sensor Networks (WSNs) are collections of sensing devices using wireless communication to exchange data. In the past decades, steep advancements in the areas of microelectronics and communication systems have driven an explosive growth in the deployment of WSNs. Novel WSN applications have penetrated multiple areas, from monitoring the structural stability of historic buildings, to tracking animals in order to understand their behavior, or monitoring humans' health. The need to convey data from increasingly complex applications in a reliable and cost-effective manner translates into stringent performance requirements for the underlying WSNs. In the frame of this thesis, we have focused on developing routing protocols for multi-hop WSNs, that significantly improve their reliability, energy consumption and latency. Acknowledging the need for application-specific trade-offs, we have split our contribution into two parts. Part 1 focuses on collection protocols, catering to applications with high reliability and energy efficiency constraints, while the protocols developed in part 2 are subject to an additional bounded latency constraint. The two mechanisms introduced in the first part, WiseNE and Rep, enable the use of composite metrics, and thus significantly improve the link estimation accuracy and transmission reliability, at an energy expense far lower than the one achieved in previous proposals. The novel beaconing scheme WiseNE enables the energy-efficient addition of the RSSI (Received Signal Strength Indication) and LQI (Link Quality Indication) metrics to the link quality estimate by decoupling the sampling and exploration periods of each mote. This decoupling allows the use of the Trickle Algorithm, a key driver of protocols' energy efficiency, in conjunction with composite metrics. WiseNE has been applied to the Triangle Metric and validated in an online deployment. The section continues by introducing Rep, a novel sampling mechanism that leverages the packet repetitions already present in low-power preamble-sampling MAC protocols in order to improve the WSN energy consumption by one order of magnitude. WiseNE, Rep and the novel PRSSI (Penalized RSSI, a combination of PRR and RSSI) composite metric have been validated in a real smart city deployment. Part 2 introduces two mechanisms that were developed in the frame of the WiseSkin project (an initiative aimed at designing highly sensitive artificial skin for human limb prostheses), and are generally applicable to the domain of cyber-physical systems. It starts with Glossy-W, a protocol that leverages the superior energy-latency trade-off of flooding schemes based on concurrent transmissions. Glossy-W ensures the stringent synchronization requirements necessary for robust flooding, irrespective of the number of motes simultaneously reporting an event. Part 2 also introduces SCS (Synchronized Channel Sampling), a novel mechanism capable of reducing the power required for periodic polling, while maintaining the event detection reliability, and enhancing the network coexistence. The testbed experiments performed show that SCS manages to reduce the energy consumption of the state-of-the-art protocol Back-to-Back Robust Flooding by over one third, while maintaining an equivalent reliability, and remaining compatible with simultaneous event detection. SCS' benefits can be extended to the entire family of state-of-the-art protocols relying on concurrent transmissions.