Recursive decentralized localization for multi-robot systems with asynchronous pairwise communication

Abstract

This paper provides a fully decentralized algorithm for collaborative localization based on the extended Kalman filter. The major challenge in decentralized collaborative localization is to track inter-robot dependencies, which is particularly difficult when sustained synchronous communication between the robots cannot be guaranteed. Current approaches suffer from the need for particular communication schemes, extensive bookkeeping of measurements, overly conservative assumptions, or the restriction to specific measurement models. This paper introduces a localization algorithm that is able to approximate the inter-robot correlations while fulfilling all of the following conditions: communication is limited to two robots that obtain a relative measurement, the algorithm is recursive in the sense that it does not require storage of measurements and each robot maintains only the latest estimate of its own pose, and it supports generic measurement models. The fact that the proposed approach can handle these particularly difficult conditions ensures that it is applicable to a wide range of multi-robot scenarios. We provide mathematical details on our approximation. Extensive experiments carried out using real-world datasets demonstrate the improved performance of our method compared with several existing approaches.

Unscented Transformation-Based Multi-Robot Collaborative Self-Localization and Distributed Target Tracking

“…Although the CL and DT problems have been extensively studied, their combination still draws limited attention, especially when considering practical multi-robot operation conditions such as nonlinear models or limited sensing and communication capabilities. One answer to the above challenge was provided in [9] by implementing the EKF scheme and asynchronized measurement update. However, this only covered the CL task, and the uncooperative target was not considered.…”

“…Motivated by the above observations, in this paper, a UT-based CLAT scheme (UT-CLAT) is proposed that can realize self-localization and target tracking simultaneously in practical multi-robot operation scenarios. The correlations are properly addressed by implementing split covariance methods, similar to the method in [9], and the covariance intersection method. The UT approach was adopted to approximate the statistics of random variables in nonlinear models.…”

“…Compared with a single robot, an MRS usually has greater efficiency and operational capability in accomplishing complex tasks, such as transportation [2], search and rescue [3], and mapping [4]. Among these MRS applications are the fundamental tasks of obtaining reliable localization information for the local robot and the uncooperative target using various measurements; these two processes are often referred to as collaborative self-localization (CL) [5][6][7][8][9] and distributed target tracking (DT) [10][11][12][13], respectively. In the CL process, each robot measures the relative quantities with regard to neighboring cooperative robots.…”

“…Then, the states of the target can be estimated cooperatively through interactions with other robots. Although the problems of CL and DT are usually solved by two separate techniques, such as in [5][6][7][8][9][10][11][13][14][15][16][17], above are consensus-based and hence may generally result in a heavy communication burden. Recently, the aforementioned DHIF was extended to a nonlinear scenario using the DT approach in [12], and the stochastic stability was analytically studied.…”

A Novel Simple Two-Robot Precise Self-Localization Method

“…The practicality of these methods is limited since they require either GPS reception which is not available in many cases such as indoor or underground areas or prior placing of sensing tools. Similarly, many solutions use filtering techniques, most commonly the Extended Kalman Filter (EKF) [9]- [12], where the robots' locations are predicted by odometry data (such as linear and angular velocities) and corrected by relative measurements between neighboring robots. Recently, the use of Ultra-Wideband (UWB) range-sensors has become popular for relative distance measurements in multi-robot systems, because they make it possible to perform the localization process in a fully decentralized manner [13]- [15].…”