The increasing demand of high data-rate services and the unstoppable growth of communicating interfaces are two important features of next-era wireless communications. Albeit some strategies have recently arisen to deal with these issues, the necessity of channel state information is still essential to guarantee the desired performance. Acquiring this channel knowledge has been traditionally solved by means of pilot reference signals, which are usually designed offline. Nonetheless, in scenarios like massive multiple-input multiple-output systems and highly dense distributed networks, it is very likely that some terminals employ the same pilot sequence due to \textit{pilot reuse}, leading to \textit{pilot contamination}. Pilot contamination is a kind of interference suffered during the side information of channel coefficients acquisition. Hence, the estimated channel may present severe inaccuracies, yielding a reduced throughput, increased outage and error probability, and imperfect secrecy performance. Therefore, a notable number of works have been devoted to studying how this interference can be combated, or even avoided. In this Master's Thesis we present an uncoordinated online pilot design scheme for opportunistic communications. It is noteworthy that this is a very demanding scenario due to the lack of coordination between users. This strategy provides pilot waveforms adapted to the scenario. In this sense, each user acquires side information of the environment (herein denoted as \textit{external network}) to design pilot waveforms which only exploit the available resources in the wireless network. Thus, the proposed methodology presents both interference management and pilot contamination avoidance capabilities. Interestingly, when the design problem is tackled from the minimum cross-interference point of view, the final design reduces to a classical minimum-norm optimization. As a first approach, we study the pilot design problem when these uncoordinated users (denoted further on as \textit{internal users}) are solely able to sense one spectral dimension. Although a very sounded study is presented accounting for the frequency dimension, an extension to the spatial dimension is also presented. This strategy is named \textit{dimension spreading}, since, ideally, internal users will spread the transmitted pilot sequence over all available degrees-of-freedom (DoF), while the occupied ones by external-network users are left unused. Dimension spreading is then extended to bidimensional scenarios, i.e. when two spectral dimensions may be simultaneously exploited. In this sense, we present an online space-frequency pilot design scheme. Internal users sense the bidimensional spectrum and exploits all frequency-angle pairs which are not occupied. Among different properties, one of the main differences with offline-designed pilots is the robustness in front of frequency calibration errors, since each internal user is self-calibrated using local side information. Furthermore, since exploiting the bidimensional spectrum may burden computational inefficiency, we present two options (which are not mutually exclusive) to tackle this problem: the asymptotic behavior of space-frequency minimum-norm pilot waveforms, and the performance of low-resolution quantization in low-SNR regimes. Finally, since the pilot are locally designed at each internal node, we prove the detectability of these references signals at other internal nodes. In this sense, we present a coordination procedure which simultaneously allows the discovering of other internal nodes, and the DoF coordination as well. Albeit the solution proposed is optimal by implementing an exhaustive search, we additionally present an efficient suboptimal two-step discovering procedure.

The next generation of wireless communications is mainly characterized by the large increase of connected devices and the high data-rates demanded to satisfy users requirements. As it is well known, the frequency spectrum is a scarce resource and, therefore, one of the identified bottlenecks to satisfy the requirements for the next era. A possible solution to overcome the aforementioned bottleneck is to account for the fact that devices will not be always transmitting data, and hence they only need the spectral resources during a certain amount of time. In this sense, diffe