Reinforcement learning (RL) is a new research area practical in the internet of things (IoT) where it addresses a broad and relevant task through about making decisions. RL enables interaction of devices and with the environment through a probabilistic approach using the response from its own actions and experiences. RL permits the machine and software agent to attain its behavior constructed on feedback from the environment. The IoTs extends to devices to the internet like smart electronic devices that can network and interconnect with others over through connectivity of remote resources being supervised and meticulous. In this paper, we examine the main four RL techniques including Markov Decision Process (MDP), Learning Automata (LA), artificial neural network (ANN), Q-learning in relation to its applicability in IoT, challenges and link them to state of art solutions. This review provides a summarized analysis of RL techniques that researchers can use to identify current bottlenecks in IoT and suggest models that are in line with the move.