This paper conducts a comparative study of three IoT-based PRNG models, including Logistic Map, Double Pendulum, and Multi-LFSR, implemented on an FPGA platform. Comparisons are made across key performance metrics like randomness, latency, power consumption, hardware resource usage, energy efficiency, scalability, and application suitability. Compared to Multi-LFSR, Logistic Map, and Double Pendulum Models provide perfect quality randomness, which is quite apt for high-security grade applications; however, the requirements of these models concerning power and hardware resources are also considerably high. By contrast, the Multi-LFSR comes into its own due to its lower latency, power consumption, and resource-efficient design. It is, therefore, suited for embedded or real-time applications. Furthermore, environmental sensors will also be introduced as entropy sources for the PRNGs to enhance the randomness of the systems, particularly in IoT-enabled battery-powered FPGA platforms. The experimental results confirm that the Multi-LFSR model has the highest energy efficiency, while the Logistic Map and Double Pendulum outperform in generating numbers with very high security. The study thus provides a deeper insight into decision-making for selecting PRNG models.