Multi-robot cardinality-balanced multi-Bernoulli filter simultaneous localization and mapping method

Real-time dense mapping technology for multi-robot systems is crucial in scenarios like search and rescue. This paper presents CMDS-SLAM, a centralized multi-robot dense surfel SLAM system aimed at overcoming limitations in hardware constraints, data transmission, and real-time creation and updating of dense maps in multi-robot SLAM. CMDS-SLAM reduces the transmission of dense information by employing a dense information ﬁltering mechanism based on co-visual keyframes, in conjunction with the extraction and compression of superpixels. Additionally, the method employs a three-stage superpixel segmentation approach to optimize transmission and enhance the eﬃciency of surfel map generation. Finally, a surfel co-visibility graph is established, and multi-robot surfel map maintenance and updates are achieved through co-visibility graph and map optimization. A comprehensive evaluation of CMDS-SLAM indicates that the method enables multi-robot surfel mapping and signiﬁcantly alleviates data transmission pressures while achieving real-time updates and maintenance of the surfel map.

Section: Introductionmentioning

confidence: 99%

CMDS-SLAM: real-time efficient centralized multi-robot dense surfel SLAM

Zuo,

Feng,

Xiao

2024

“…Comparatively, the second class, which is based on random finite set (RFS) theory, can greatly mitigate computational complexity. The RFS-based filters include probability hypothesis density (PHD) filter [13], cardinalized PHD filter [14], cardinality-balanced multi-Bernoulli filter [15], and Poisson multi-Bernoulli mixture filter [16]. Under the framework of RFS, PHD filter, which propagates posterior intensity (i.e.…”

Section: Introductionmentioning

confidence: 99%

Underwater multi-source DOA tracking using uniform linear array based on improved GM-PHD filter

Zhang

Liang

et al. 2023

Uniform linear array (ULA) is commonly used for collecting underwater acoustic signals, and the direction-of-arrival (DOA) is obtained by array signal processing methods. The ULA-based multi-source DOA tracking suffers from the performance degradation caused by the presence of false tracks from port-starboard ambiguity. To address this problem, we propose a bearing ambiguity discrimination algorithm. Firstly, the sum of real bearing motion velocity between truth and ambiguous targets is established with the Gaussian distribution. Secondly, an improved Gaussian mixture probability hypothesis density (GM-PHD) filter, which can adaptively generate target birth intensity, is utilized to select truth and ambiguous bearings generated by the same target from the estimated posterior intensity. Thirdly, truth and ambiguous bearings are discriminated according to the fact that the change of array heading will lead to a significant change in ambiguous bearing. Simulation and real experimental results illustrate that the proposed algorithm can provide unambiguous target-generated tracks and accurately estimate the number of targets.

“…This led to Vo et al [19] proving the existence of the cardinality overestimation problem in the MeMBer filter, thus further proposing the cardinality balanced MeMBer (CBMeMBer) filter and its implementation of GM and SMC. In some applications [20][21][22], the CBMeMBer filter can extract and estimate states more conveniently and reliably than the traditional PHD filter.…”

Section: Introductionmentioning

confidence: 99%

An interval SLAM algorithm for leg–arm mobile robots based on a CBMeMBer filter with Gaussian indicator box-particles

Luo

Qin

2023

During the implementation of the SLAM methods based on probability hypothesis density (PHD) with sequential Monte Carlo (SMC), it is necessary to generate a relatively large number of particles for mapping and cluster particles for target state extraction, which brings about heavy computing requirements and difficulty in enhancing the estimation accuracy. Along these lines, in this work, an interval SLAM algorithm for Leg-Arm mobile robot (LA-MR) was presented, which was based on cardinality-balanced multi-target multi-Bernoulli filter with Gaussian indicator box-particle (GIBP-CBMeMBer). First, an effective strategy for improving the selection criterion of subdivision box-particle was given via the indicator subdivision resampling (ISD-resampling) to provide more reliable box-particles. In addition, mixtures of Gauss distributions with box supports were developed to fit posterior probability, i.e. Gaussian indicator box-particle (GIBP). Then, the GIBP filter was employed to implement cardinality balanced multi-target multi-Bernoulli (CBMeMBer) filter and thus the particle numbers were significantly reduced to decrease the computational complexity compared with the SMC-CBMeMBer filter. In this way, the feature map learning was considered as a multi-target tracking problem, whereas a kind of SLAM algorithm for LA-MR was also designed and implemented through the GIBP-CBMeMBer filter, in which the GIBP filter was used to realize simultaneous localization and the GIBP-CBMeMBer filter was utilized to build a map. A series of experiments and a case study clearly illustrated the performance superiority of the proposed algorithms.