Because of the advances of technology, the number of devices connected to the Internet now exceeds the number of people on Earth. These devices are not only computers, tablets and smartphones, but also any object that has an embedded microcontroller and a radio with an antenna. Until recent years, the Internet was an interconnection of computer networks. We are talking now about networks of computers and Things. New protocols have been standardized in order to integrate these networks in the Internet. Among them, the IEEE 802.15.4 physical and MAC layer protocols, and RPL, the IPv6 Routing Protocol for Low-power and Lossy Networks. This resulted in a new paradigm, called the Internet of Things. An important part of the Internet of Things are the Wireless Sensor Networks (WSNs), which are composed of a large number of low-power devices that sense the environment and communicate this information to a central entity. The WSNs are highly unreliable, and the devices that compose them are very constrained in terms of energy, memory, and processing power. The goal of this thesis is to improve the already standardized protocols for the Internet of Things, considering the constraints of WSNs (high loss rate, instability, low data rates, etc.), and of the devices that compose it. First, we proposed a new MAC layer broadcast mechanism in IEEE 802.15.4, to ensure a reliable delivery of the control packets from the upper layers (especially from RPL). Then, we provided an exhaustive evaluation of RPL, and highlighted an instability problem: existing routing metrics used to build the topology never lead to network stability. These instabilities generate a large overhead, consuming a lot of energy. Since the lifetime of WSNs is very limited, as most of their devices are energy constrained, we proposed a new routing metric that estimates the Expected Lifetime of a node, according to its residual energy, the link quality to its neighbors, and the current traffic conditions. By using this metric with RPL, we identify the bottlenecks of the network in terms of energy (i.e., the nodes that have the least residual energy), and we route the packets in order to maximize their lifetime. Finally, we proposed to enhance RPL by proposing a multipath approach, in order to create energy-balanced paths. We identify the weakest nodes in the network (in terms of energy) by using the Expected Lifetime metric. Then, we forward the traffic through all the next hops, so that each path consumes the same quantity of energy, maximizing furthermore the lifetime of the network.