Research On Simultaneous Localization And Mapping Technology Oriented To Dynamic Environment

It is relatively difficult to estimate the accurate pose of the sensor platform continuously when GPS or other absolute positioning is not available.Simultaneous Localization and Mapping(SLAM)is the core technology to solve this problem.Visual Inertial Navigation System(VINS)can achieve high accuracy,so it is currently a popular research field.But for the sensor platform,such as a ground vehicle,the additional unobservable direction,scale,will be introduced into VINS when it moves on a plane at a constant speed or acceleration.Since these movements occur frequently,the positioning accuracy will also be affected.In addition,accelerometer measurements can be greatly affected by the noise.Therefore,a better bias estimation of accelerometer is needed for an accurate integration result.In a dynamic environment,when moving objects occupy a small field of view(such as moving objects in the distance),removing unreliable features from moving objects can achieve certain results.However,when the moving object occupies a large field of view(such as a moving object in the vicinity),multi-sensor fusion can be considered to improve the robustness of the algorithm.To deal with these problems,the SLAM algorithm in this paper based on the fusion of vision,gyroscope and wheel encoder is researched and implemented for the ground vehicle that transports goods in the dynamic indoor warehouse environment.The main contents include:(1)Establish pre-integration model for the wheel encoder.Compared with the accelerometer,the wheel encoder can provide more reliable translation information and overcome the additional unobservability of scale.(2)Merge visual constraints,gyroscope constraints,wheel encoder constraints and plane constraints into a unified back-end optimization framework.(3)Use geometric methods to remove outliers of visual features.And when the visual feature tracking fails,the gyroscope and wheel encoder are used to provide reliable pose estimation in a short time.(4)Analyzes the external parameters between the wheel encoder and the camera,and uses the analysis results to guide the online optimization of the external parameters.Considering the computing performance of the hardware platform and the actual application of the ground vehicle,the SLAM algorithm in this paper does not include the loop closure detection module.This paper proposes a solution which is fusion of vision,gyroscope and wheel encoder.The complementary advantages provide a high-frequency and high-precision odometer for the ground vehicle.Experimental results show that the algorithm implemented in this paper has good positioning performance and computing performance.