Biomorphic navigation system version

Abstract

The purpose of this work is to create and study the dynamics of the functioning of a biorelevant visual navigation system. Methods. The work uses simultaneous navigation and mapping systems RatSLAM and Orb-SLAM. The RatSLAM system is a biorelevant model of visual navigation in the rodent hippocampus. The Orb-SLAM system is a simultaneous navigation and mapping system that works on the principle of searching and tracking changes in the position of key points in the image. Results. The article presents a version of a modified visual navigation system. The system consists of a visual odometry module based on the Orb-SLAM system, as well as a mapping and loop closure module based on the RatSLAM system. This allows you to combine the localization accuracy of systems operating on the principle of tracking key points in the image and neural filtering of biorelevant systems. Using the constructed system, location estimates were obtained on public and new data sets. Conclusion. The constructed visual navigation system determines the location of the subject (video camera) in space, which is in good agreement with the ground truth location data.

Keywords

• simultaneous localization and mapping systems
• path integration
• landmarks
• neural networks