Mobile network operators have historically experienced increasing traffic loads at a steady pace, which has always strained the available network capacity and claimed constantly for new methods to increase the network capacity. A key solution proposed in the context of mobile communications to increase the available spectrum is the exploitation of the unlicensed spectrum in the 5 GHz ISM bands, predominantly occupied by Wi-Fi systems based on the popular IEEE 802.11 standards family. However, given the medium access protocol implemented by IEEE 802.11 networks, an uncontrolled deployment of mobile networks in unlicensed bands could potentially lead to a resource starvation problem for Wi-Fi networks and therefore degrade their performance significantly. To address this issue and allow a fair coexistence of mobile and Wi-Fi networks in the unlicensed spectrum, the 3rd Generation Partnership Project (3GPP) standardised the Long-Term Evolution Unlicensed (LTE-U) and Licensed Assisted Access (LAA) technologies. The main philosophy of these technologies is to allow mobile operators benefit from the vast amount of available spectrum in unlicensed bands without causing a significant performance degradation to existing Wi-Fi networks, thus enabling a fair coexistence. However, the proposed coexistence mechanisms at both the physical and (mainly) MAC layers have been proven to provide very limited guarantees of fairness, if any at all. This thesis investigates and proposes several improvements to the 3GPP coexistence mechanisms to enable a truly fair coexistence between mobile and Wi-Fi networks in unlicensed bands. In particular, various methods are proposed to adjust the transmission duty cycle in LTE-U and to adapt/select both the waiting and transmission times for LAA. The main novelty of this work is that the proposed methods exploit the knowledge of the existing Wi-Fi activity statistics to tune the operating parameters of the coexistence protocol (duty cycle, contention window size and its adaptation, transmission opportunity times, etc.), optimise the fairness of spectrum coexistence and the performance of mobile networks. While Wi-Fi activity statistics can be readily obtained in commercial products, few solutions actually exploit them. This research shows that, by means of a smart exploitation of the knowledge of the Wi-Fi activity statistics, it is possible to guarantee a truly fair coexistence between mobile and Wi-Fi systems in unlicensed bands. Compared to the 3GPP coexistence mechanisms, the proposed methods can attain a significantly better throughput performance for the mobile network while guaranteeing a fair coexistence with the Wi-Fi network. In some cases, the proposed methods are able not only to avoid degradation to the Wi-Fi network but even improve its performance (compared to a coexistence scenario between Wi-Fi networks only) as a result of the smart coexistence mechanisms proposed in this thesis. The proposed methods are evaluated for the 4G LTE standard but are similarly applicable to other more recent mobile technologies such as 5G NR-U​​