A Brain-like Cognitive Decision-making Model Based On The Regulation Mechanism Of Cerebellum And Basal Ganglia

At present,artificial intelligence provides a powerful driving force for the scientific and technological revolution in the new era.The traditional artificial intelligence represented by machine learning has been applied to many research fields.However,machine learning often faces many difficulties,such as the need for a large number of training data.In order to break through this bottleneck,a large number of researches began to turn to the exploration and imitation of brain physiological mechanism.More and more researchers think that brain like intelligence will be the development direction of artificial intelligence in the future.Brain like intelligence also gets more and more attention and research by virtue of low power consumption,weak supervision and other advantages.Robot behavior decision algorithm or behavior control strategy is the research focus in mobile robotics.In this paper,we will explore and simulate the physiological structure related to behavior decision in the brain,and then apply it to the behavior decision and autonomous navigation of mobile robots.Firstly,based on a large number of brain-like biomimetic algorithms related to goal orientation and decision-making,combined with the physiological mechanism of cerebellum and basal ganglia,a brain like cognitive decision-making model based on the regulation mechanism of cerebellum and basal ganglia is proposed.The model simulates the functions of basal ganglia and cerebellum in human brain,and connects them to form a hybrid model which can realize the behavior decision of mobile robot.The hybrid model uses the MDN to simulate the cerebellum,so as to speed up the learning convergence of the basal ganglia;at the same time,through the reinforcement learning based on the Actor-Critic to simulate the basal ganglia,so as to update and improve the knowledge base of the cerebellum model,so that the cerebellum can obtain better decision-making ability in the subsequent behavior.Furthermore,the action selection strategy of the basal ganglia in the hybrid model was changed to the action selection strategy based on ACC(anterior cingulate cortex).This method is based on the phase mode and tonic mode of the anterior cingulate cortex and prefrontal cortex.Through the simulation of anterior cingulate cortex and related neural tissue structure and the introduction of alertness and other related parameters,the agent can enhance the exploration behavior when receiving negative feedback,and enhance the utilization behavior when receiving positive feedback,so as to achieve the dynamic balance between the exploration environment and the utilization environment,and further improve the behavior decision quality of the hybrid model.Finally,after the completion of the model construction,the hybrid model is applied to the autonomous navigation scene of mobile robot,and the effectiveness of the hybrid model in behavioral decision-making and autonomous navigation is verified by a variety of experimental methods.Firstly,this paper applies the hybrid model to the simulation experiment of robot autonomous navigation based on MATLAB platform;then,the hybrid model is compared with a variety of widely used algorithm models;finally,the hybrid model is transplanted to mobile robot based the ROS,The autonomous navigation experiments are carried out on the gazebo physical simulation platform and the real Rikirobot platform.The experimental results show that the model has the advantages of faster learning convergence,better effect after convergence and higher adaptability in robot behavior decision-making and autonomous navigation.