Efficient Decentralized Collaborative Mapping for Outdoor Environments

Abstract

Abstract-An efficient mapping in mobile robotics may involve the participation of several agents. In this context, this article presents a framework for collaborative mapping applied to outdoor environments considering a decentralized approach. The mapping approach uses range measurements from a 3D lidar moving in six degrees of freedom. For that case, each robot performs a local SLAM. The maps are then merged when communication is available between the mobile units. This allows building a global map and to improve the state estimation of each agent. Experimental results are presented, where partial maps of the same environment are aligned and merged coherently in spite of the noise from the lidar measurement.

On Data Sharing Strategy for Decentralized Collaborative Visual-Inertial Simultaneous Localization And Mapping

“…Schuster et al [4] proposed a submap-based approach for visual-stereo SLAM that makes each robot share some local submaps with neighboring robots, which are then aligned via ICP point-cloud registration. Contreras et al [5] also proposed a decentralized method for LiDAR SLAM. To cope with limited bandwidth constraints, other works make robots exchange summarized representations of their map to meet the bandwidth requirements and ease their alignment within the recipient robot's map.…”

“…The method proposed in this section is based on the selection of landmarks and raw measurements to compute data packets that will be exchanged between robots. An overview of the MKCS-VIP method is given by the figure (5). All the process mentioned in this figure are detailed below.…”

CoMapping: Multi-robot Sharing and Generation of 3D-Maps applied to rural and urban scenarios

“…Several methods have been proposed to merge maps. In [2] and [11] techniques were proposed for 3D merging of occupancy grid maps based on octrees [12]. Their merging process refined the transformation estimate between maps by ICP registration [8].…”

“…Their merging process refined the transformation estimate between maps by ICP registration [8]. Specifically, in [11] an ICP version was performed including an efficient technique to exchange maps between robots in order to optimize the bandwidth resources of a multi-robot network for decentralized cases. Those last cases are experimentally studied in this paper.…”

“…In the following sections, our metrics from (3) will be renamed asε r * andε t * , associated to experiments with unknown maps transformations. In the Pre-Local Mapping stage, inspired by the work in [11], its versatile configuration does not depend on a specific LidarSLAM method. A modified version of 1 [10] was used as LidarSLAM method for this article, considering its good place in the KITTI ranking 2 .…”