Cloud infrastructure has enabled novel business models leading to exponential growth of cloud usage in diversified applications. Consumers of cloud offerings include individuals as well as organizations of different sizes. Therefore, the workloads across the users are very dynamic in nature. Workloads consist of numerous jobs or tasks to be executed in cloud infrastructure. Due to dynamic nature of workloads, the cloud infrastructure usage and availability of free computing resources are also dynamic. In this context, the traditional heuristics-based algorithms are not capable of understanding runtime scenarios and fail to provide optimal scheduling benefits. Therefore, it is important to explore artificial intelligence (AI)-enabled approaches like deep learning (DL) and machine learning (ML). From the systematic review of existing methods that are based on machine learning, it is understood that reinforcement learning (RL), a deep learning model, is suitable for improving performance in task scheduling. The rationale behind this is that RL is based on an agent making decisions based on the runtime state space, action space and feedback gained. This paper covers review of state-of-the-art methods from 2011 to 2023. This paper imparts insights to the reader regarding various facets of task scheduling in cloud environments, thereby opening the door to potential avenues for further research aimed at achieving more refined task scheduling models.