Although machine Reinforcement Learning (RL) was developed with direct intuition from animal learning, it was not initially intended to explain the neurobiological processes of the brain. But the collaboration of neuroscientists, roboticists and computer scientists has recently permitted to draw parallels between the computational steps of RL algorithms and the neurobiological processes. This system equivalence became even more evident after the invention of recurrent neural networks (RNNs) like LSTM (Long Short-Term Memory) and Transformer networks, which seem to fulfill the function of short-term memory in the brain’s hippocampus. In this paper, we outline the relevance of RL in cognitive behavior especially in the context of action selection, spatial memory and navigational strategies. We are supporting our conclusions by mathematical models of the involved brain regions and the interplay of the neuromodulators that are engaged in these cognitive tasks. We also present the recent findings about Meta-Learning as the leading RL class of algorithms which came as the most tenable solution to the problem of sample inefficiency in generic RL. Therefore, Meta-Learning and Meta-RL models offer the closest match and the simplest explanation for the fitness of humans in performing the sophisticated cognitive and navigational tasks.