RatSLAM Optimized For Path Planning Brain-inspired Navigation

Mobile robots achieve a given task through environmental awareness and decision-making.Their effectiveness directly depends on their autonomous navigation ability in completing the task.The complex environment poses more severe challenges to the intelligent navigation model of mobile robots.Although important progress has been made in artificial intelligence,there is still a long way to go in the exploration of intelligent navigation and path planning.Therefore,basing on the cognitive,learning,and judgment abilities of mammals in the hippocampus and aiming at the limitations of the RatSLAM,this dissertation proposes an experience map construction and path planning method that combines brain-inspired cognitive theory,computer modeling,and episodic memory,which provides a new idea for brain-inspired navigation.Firstly,this dissertation constructs a brain-inspired bionic navigation model based on RatSLAM.Local view cells are created using the Scanlines Intensity,a visual processing method,which includes environmental information and simulates the processing process of human beings on external environmental information;Representing information about the motion of mobile robot through the Head-direction cells and the Place cells;Selecting and integrating self-motion information through Continuous Attractor Neural Network model to form a positional representation of the mobile robot in the environment.The perception of the environment by robots is the first step in information processing.Aiming at the problems of poor robustness in the original visual processing methods,such as being completely dependent on pixel intensity,being susceptible to illumination intensity,and having many mismatches in Closed-loop detection,this dissertation proposes to use FrequencyTuning algorithm to highlight important information in images firstly and use SURF feature matching algorithm to obtain the direction and position information of the mobile robot in the environment.Navigation accuracy and robustness are greatly improved.Inspired by the mammalian hippocampal episodic memory system,a episodic memory model was constructed which can implement path planning for mobile robots in experience maps.At the same time,a way that combination of road detection algorithms and episodic memory is proposed to detect and judge potential safe paths.Combining the detected safe paths with experience maps can achieve target-oriented optimization path planning for mobile robots.In order to verify the feasibility and effectiveness of the method in this dissertation,multiple sets of experiments were conducted in simple,complex,indoor,outdoor,and other environments.Experiments show that the proposed method has the highest accuracy compared to traditional RatSLAM and SURF-RatSLAM models;Compared with the episodic memory path planning model without road detection algorithm,the proposed method in this dissertation has the best path planning results.In summary,the algorithm in this dissertation achieves the expected requirements in navigation and intelligence.