Research on Dynamic Task Scheduling in Cloud Computing focuses on developing adaptive methods to optimize resource allocation, improve execution efficiency, and handle real-time variability in cloud environments. Recent studies highlight AI-driven scheduling models that address challenges such as resource heterogeneity, energy efficiency, and multi-cloud adaptability. Hybrid optimization approaches, including combinations of Particle Swarm Optimization with Grey Wolf Optimization and Transformer-based models with Cuckoo Search, enhance convergence speed and scheduling quality. Deep Reinforcement Learning methods provide intelligent job scheduling and resource management, while predictive graph networks support dynamic load balancing in heterogeneous cloud systems. Collectively, these advancements enable more responsive, scalable, and efficient dynamic task scheduling frameworks for modern cloud computing.