A Possible Explanation for the Generation of Habit in Navigation: a Striatal Behavioral Learning Model

Efficient behavioral learning is a great challenge to an autonomous mobile robot. During navigation, animals can improve their behavioral learning ability by constantly interacting with the environment and gradually realize efficient navigation. Although habitual behavior in animal navigation is relatively well known, understanding of the brain’s habit-generation mechanism remains limited. In this study, we propose a striatal behavioral learning model, composed of the striosome and a matrix model, to possibly explain the generation of habitual behavior in animal navigation. The model’s bionic mechanism is characterized as follows: (1) in the striosome model, orientation information updates constantly based on the operant conditioning mechanism, leading to the generation of habitual behavior, and (2) a matrix model with an improved e-greedy algorithm chooses actions by adjusting the utilization of learned habits, balancing the relationship between exploration and exploitation in an agent’s navigation. We test our model in Morris square dry maze tasks. Results indicate the effectiveness of the model in explaining habit-related behavior. Besides, we compare our model with the widely used striatal temporal difference learning model. Results show that our model is more efficient and robust than the contrast model. We can conclude that it can successfully solve navigation tasks with habits while showing key neural characteristics of the striatum, which may be significant to the bionic navigation of robots. The proposed model confirms and builds a relationship among habit generation, the striatum, and operant conditioning, which may help explain the mechanism underlying habit generation in animal navigation.