Abstract
This paper proposes a novel distributed EKF-SLAM system that combines the advantages of EKF-SLAM and distributed SLAM systems. The system model of this novel SLAM system has a distributed structure, and each subsystem is a special SLAM system corresponding to every effectively observed landmark by feeding the heading information from a magnetic compass is introduced into the observation equation. Aim at the correlation problem in distributed SLAM system, a decorrelated distributed EKF (DDEKF) was developed to estimate the robot pose and landmarks. DDEKF reconstructs and extends the maximum allocation covariance (MAC) method so that it can be applied to the distributed structure where the number of local filters is dynamically changed. Then, the local filter estimation results are weighted and fused in the main filter to obtain the estimation result. Finally, the experimental tests were performed in an outdoor environment, and the experiment results demonstrate that the proposed novel distributed EKF-SLAM system has a better performance than the existing SLAM system.
Highlights
Simultaneous localization and mapping, a well-known computational problem abbreviated as SLAM, was proposed as a means of enabling a mobile robot to move through an unknown environment while building a map and simultaneously estimating its position [1]
In order to evaluate the performance of the proposed distributed structure distributed extended Kalman filter (EKF) (DDEKF)-SLAM system and consider the problem of correlation between local filters to the estimation accuracy, a robot runs in the square to collect the actual data, and the DDEKF algorithm is used to process it
This paper has proposed a decorrelated distributed SLAM system based on the EKF, which is an effective real-time solution for the automation navigation of mobile robots
Summary
Simultaneous localization and mapping, a well-known computational problem abbreviated as SLAM, was proposed as a means of enabling a mobile robot to move through an unknown environment while building a map and simultaneously estimating its position [1]. Any change in the number of feature points will result in state vector reconstruction and an additional computational burden because all of the state information is contained in one vector, and information of different quality will be unavoidably confused To overcome these limitations of centralized filters, Dae Hee Won proposed a SLAM system based on a distributed particle filter (DPF) [20, 21]. Since EKF estimation directly provides recursive solutions to localization problems and suitable to the robot and landmark positions, the EKF approach remains the method of choice for the great majority of applications, and EKFSLAM has been proven to offer the best convergence and consistency [21, 22] Motivated by this previous experience, this paper proposes a decorrelated distributed EKF-SLAM system.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
