The development of processors and sensor technologies and the advent of artificial intelligence (AI) have led to the growth of more capable and affordable mobile intelligent agents like unmanned aerial vehicles (UAVs) which have numerous applications in areas such as construction, search and rescue operations, oil and gas, agriculture, etc. The use of AI and machine learning (ML) algorithms in UAVs has enabled real-time data-driven decision-making frameworks but poses numerous challenges including high-impact decision-making, security, energy management, communication, and automation, to name a few. In this dissertation, we propose solutions to some of these problems. First, we give our attention to the probabilistic decision-making problem with an emphasis on the problems where the law of large numbers (LLN) does not apply. These problems are referred to as ``high-impact" problems. We present a comprehensive decision-making framework based on change-of-probability measures that addresses both the high-impact and ordinary (non-high-impact) problems, offering a systematic approach for rational decision-making. In other words, the proposed framework addresses the lack of concrete frameworks for rational decision-making, especially in the high-impact scenario. Subsequently, we move on to discuss the privacy issue, with particular attention paid to location privacy, the lack of which presents significant risks for users. In this regard, we first propose a model-free privacy-preserving mechanism while imposing a continuity constraint on the location data to avoid leakages from abrupt jumps. We introduce an algorithm where the privacy and utility guarantees as well as the continuity constraint are met. The location privacy problem is then investigated in UAV applications, in particular, UAV package delivery and the Internet of Things (IoT) data collection. The goal is to protect the location privacy of a UAV destination's location privacy from an adversary observing its trajectory. Considering performance metrics corresponding to each application, we analyze the privacy-utility tradeoffs. In particular, we provide numerical results for the tradeoff between privacy and mean peak age of information (PAoI) in the IoT data collection application.