Bionic SLAM based on MEMS pose measurement module and RTAB-Map closed loop detection algorithm

Abstract

In the complex indoor scene, the RatSLAM, a rodent model navigation algorithm, will suffer a performance reduction due to light changes or other factors. Based on this the RTAB-Map closed loop detection strategy is introduced into the RatSLAM system, which can, through closed loop detection, eliminate the accumulative errors cause by the experience map of pose cells and local view cells, and thus to improve the instability of performance due to light changes or other factors. However complex scene, such as moving obstructions, will lead to mistakes in the visual odometer’s identification of speeds and thus cause conspicuous skewing of the navigation trail, which sometimes cannot be corrected through scene reorientation. This paper proposes a RatSLAM model with pose measurement module and RTAB-Map closed loop detection algorithm. The improvements are as follows. First, by fusing RTAB-Map closed loop detection, the phenomenon of odometer drifting under RatSLAM bionic algorithm due to error accumulation of friction or other factors like light changes can be improved; second, RTAB-Map algorithm itself improves the system real-time performance by using four kinds of memorizers; last but not least, the fusion of pose measuring module can prevent emergent obstruction from disturbing the visual odometer to obtain speed information, and it is more accurate when combined with sensor technology.