UC Berkeley

The density and frequency of interaction between society and technology is increasing and this presents us with opportunities to improve lives and livelihoods, and also to address the inequities in our society holistically. The Internet of Things (IoT) and 5G wireless com- munication technologies are the future technologies that envision to bring us closer to these opportunities. These technologies leverage the ubiquitous sensing, actuation, and comput- ing capabilities to enable smart devices to perform interesting tasks and to gain knowledge about the environment. These may include futuristic healthcare systems, affordable preci- sion agriculture systems, smart energy-efficient cities, and advanced flexible manufacturing. However, the present understanding of wireless communication is not enough to get us to this future.

In this dissertation, we look at designing wireless communication frameworks for interac- tive applications like drone swarms, and industrial automation that require fast and reliable communication. We focus on designing frameworks that make wireless communication highly reliable while also maintaining the latency requirements of the systems being targeted. This is the missing link needed to leverage the power of wireless communication for the next generation of interactive applications.

One of the key contributions of this thesis is the design of cooperative communication based protocol frameworks that leverage a combination of diversity techniques to achieve the target reliability and latency. The framework uses spatial diversity to combat multipath channel fading and repetitions in time and frequency to shield against unmodeled error events. We analyze these protocols using the communication-theoretic delay-limited-capacity framework and consider their sensitivity to different modeling assumptions.

Another key contribution of this thesis is an in-depth exploration of the dynamics of wireless channels in the context of ultra-reliable low-latency communication (URLLC). We revisit some standard concepts such as coherence time and question whether some of the modeling assumptions made in the context of cellular and WLAN communications make sense in the context of URLLC. We find that our cooperative communication frameworks are robust to the nominal dynamic channel models, especially spatial dependence. However, events such as synchronization mismatch or sudden change in the channels due to shadow- causing objects need to be protected against and therefore we build in frequency and time margins.

The final contribution of this thesis is bringing together the temporal model of channel dynamics and machine learning to build intelligent relay selection strategies. This essentially provides the reliability needed by smartly selecting a small set of relays instead of relying on using every single relay available to combat fading. Finally, we present some preliminary experimental findings that pave the way to making these systems practical.