Research On Improved RatSLAM Algorithm Of Fusing GDSOM

SLAM(Simultaneous Localization and Mapping),which is a synchronous positioning and map construction,can be described as a mobile robot moving in an unknown environment,not only needs to establish the environment map according to the known robot pose,but also needs to determine the new pose of the robot based on the obtained environment map.Based on the current research status of SLAM and summarizing the characteristics of various SLAM algorithms,this paper selects the cognitive and neural-inspired RatSLAM model as the research algorithm of this paper.Through analysing of the characteristics of the RatSLAM model deeply,it is found that in the long-range autonomous navigation process of mobile robots,due to the influence of external conditions such as light and weather,the image information captured by the RGB camera has undergone a large change,the accuracy of matching relying on the collected image information of the RGB camera is not high,which easily causes the success rate and accuracy of the closed-loop detection algorithm to decrease.This paper applies the RGB-D information of the scene to the RatSLAM system by proposing the depth camera Kinect to replace the RGB camera.By introducing depth information and combining the original RGB image information,a new visual template is constructed to improve the closed-loop detection process of the algorithm,which reduces the mismatch rate in the closed-loop detection process and improves the positioning accuracy.At the same time,in the indoor structured environment(single environmental characteristics)of the RatSLAM algorithm,there are few environmental features that rely on visual acquisition,and there are problems of large positioning and mapping errors.A two-dimensional dynamic self-organization map(GDSOM)has proposed.By introducing environmental features and directional features and designing the robot to move along the wall to realize the self-organizing extraction of environmental road signs,the extracted environmental road sign information will be storage together with the image information as the scene feature template,the scene feature template contains more environmental information,which makes the positioning and mapping of the robot under the RatSLAM algorithm more accurate.The two improved algorithms for the RatSLAM model proposed in this paper are verified by Matlab.The simulation results show that the introduction of depth information effectively improves the shortcomings of the traditional RatSLAM model on light sensitivity,significantly improves the accuracy of closed-loop detection matching,and makes the constructed experience map more accurate.The proposed two-dimensional GDSOM network algorithm is effective to extract structural environment road sign.By constructing a new environmental feature template,improved the problem that the RatSLAM model has large error when relying on few features visual information for closed-loop detection in a structured environment,and the positioning and mapping capabilities of RatSLAM model has improved in the structured environment.