To achieve spatial cognition and autonomous navigation for the robot, learning from the biological mechanism for navigation, a goal-oriented navigation model based on multi-scale place cells map is proposed. First, a place cell map with different scale is constructed for coding the space environment. Second, the Gaussian model is used to simulate the firing rate of a place cell which will be the input of the Q-learning process. Third, the annealing strategy is explored to choose a reasonable

more »

... ion and update the Q table. After training and learning, the robot could succeed to finish the goal-oriented navigation task. Simulation results show that, the proposed method is feasible to complete the goal-oriented navigation task. Compared with the spatial cognitive model of single scale place cells, the method not only meets the multi-scale spatial representation nature of place cells in hippocampus, but also has a faster learning speed. The achievement of this paper has a reference for the study of brain-inspired navigation.