A Goal-Directed Behavioral Learning Model Based on Hippocampal-Striatal Circuit

In order to make mobile robot more bionic and intelligent, a goal-directed behavioral learning model based on hippocampal-striatal circuit (GBL-HSC) is proposed, which attempts to explore the physiological mechanisms of goal-directed navigation. Hippocampal-striatal neural circuit (HP-Str circuit) is increasingly being considered as an appropriate model system to understand goal-directed decision-making and planning mechanisms in the brain. Hippocampus provides a place cells-based cognitive map of the environment, and makes behavioral decisions according to a sequence planning algorithm of place cells. Striatum includes the striosome and the matrix, with the former receives dopamine reward signal from substantia nigra and outputs evaluation information of the behavior, which is used by the matrix to choose actions. The agent chooses action at decision point based on confidence in the prefrontal cortex. Simulation experiments of navigation in Tolman detour maze is conducted. Results show that the model can help agent to learn the optimal behavioral decision, to quickly adapt to changes in the environment, such as choosing a detour path, thus realizing a better and faster navigation.