The Robot Reature-Coding And Feature Mapping And Location Mothed Study

The order of line segments feature data accesses in memory is chaotic after the robotdetecting some areas in the process of robot SLAM. The difficulty of SLAM is how to revertto an accuracy map quickly using the data in the memory. The paper adopts line segmentsfeature coding to solve this problem. It judges the lines whether they belong to the sameobstacles according to the connected relation of lines. Then it marks and sorts line segmentsaccording to their proximity relation. This method can save the mapping time of SLAM. Ituses straight line multimetering model achieving the judgment of dots and segments in orderto remedy the ultrasonic sensor which has the weakness of sparsity. It improves the matchingspeed and accuracy of the map when the local map is matching to the segment features of theglobal map in the process of location. The EKF algorithm is used in robot SLAM of manydifferent environments by a lot of researchers. But it can’t meet the real-time mapping inover-all situation because of the calculation complex of the EKF algorithm. Moreover it’seasy to appear the local linear hypothesis when the EKF is used in larger non-linear system. Itmay lead to EKF algorithm divergence if higher order term error is not considered. So theurgent problem is to find better probability estimate algorithm to solve SLAM. This paperputs forward three better algorithms based on volume Kalman filtering and particle filterSLAM algorithm. Iterated Square-root Cubature Kalman Filter based SLAMalgorithm(ISRCKF-SLAM) of the first kind. The main contribution of the algorithm is thatthe numerical integration method based on cubature rule is directly used to calculate theSLAM posterior probability density. To improve innovation covariance and cross-covariance,the latest measurements are iteratively used in the measurement update. The algorithm canreduce linearization error and improve the accuracy of the SLAM algorithm.Adaptive IteratedSquare-Root Cubature Kalman Filter based SLAM algorithm(AISRCKF-SLAM) of thesecond kind. The algorithm also used adaptive iterating estimation restricted by the iterativesentencing guideline to adjust the proportion of the observation and state model, to make theestimated square root of the error covariance more accurate and reasonable based on the basisof the ISRCKF. It can improve estimation precision, stability and convergence. The squarecubature particle filter simultaneous localization and mapping (SRCPF-SLAM) of the thirdkind. The algorithm fuses the latest measurement information in the stage of the priordistribution updated of the particle filter SLAM. It designs importance density function by theSRCKF that is more close to the posterior density, and it spreads the square root of statecovariance. So, the algorithm ensures the symmetry and the positive semi-definiteness of thecovariance matrix and improves numerical estimation precision and stability.