Machine-to-machine (M2M) communication is becoming an important aspect of warehouse management, remote control, robotics, traffic control, supply chain management, fleet management and telemedicine. M2M is expected to become a significant portion of the Industrial Internet and, more broadly, the Internet of Things (IoT). The environments in which M2M systems are expected to operate may be challenging in terms of radio wave propagation due to their cluttered, multipath nature, which can cause deep signal fades and signal depolarization. Polarization diversity in two dimensions is a well-known technique to mitigate such fades. But in the presence of reflectors and retarders where multipath components arrive from any direction, we find the detrimental effects to be three-dimensional and thus consider herein mitigation approaches that are also 3D. The objectives of this dissertation are three. First, to provide a theoretical framework for depolarization in three dimensions. Second, to prepare a tripolar antenna design that meets cost, power consumption, and simplicity requirements of M2M applications and that can mitigate the expected channel effects. Finally, to develop new channel models in three dimensional space for wireless systems. Accordingly, this dissertation presents a complete description of 3D electromagnetic fields, in terms of their polarization characteristics and confirms the advantage of employing tripolar antennas in multipath conditions. Furthermore, the experimental results illustrate that highly variable depolarization occurs across all three spatial dimensions and is dependent on small changes in frequency and space. Motivated by these empirical results, we worked with a collaborating institution to develop a three-dimensional tripolar antenna that can be integrated with a commercially available wireless sensor. This dissertation presents the testing results that show that this design significantly improves channels over traditional 2D approaches. The implications of tripolar antenna integration on M2M systems include reduction in energy use, longer wireless communication link distances, and/or greater link reliability. Similar results are shown for a planar antenna design that enables four different polarization configurations. Finally, the work presents a novel three-dimensional geometry-based stochastic channel model that builds the channel as a sum of shell-like sub-regions, where each sub-region consists of groups of multipath components. The model is validated with empirical data to show the approach may be used for system analyses in indoor environments.