Voxgraph: Globally Consistent, Volumetric Mapping Using Signed Distance Function Submaps

Abstract

Globally consistent dense maps are a key requirement for long-term robot navigation in complex environments. While previous works have addressed the challenges of dense mapping and global consistency, most require more computational resources than may be available on-board small robots. We propose a framework that creates globally consistent volumetric maps on a CPU and is lightweight enough to run on computationally constrained platforms. Our approach represents the environment as a collection of overlapping Signed Distance Function (SDF) submaps, and maintains global consistency by computing an optimal alignment of the submap collection. By exploiting the underlying SDF representation, we generate correspondencefree constraints between submap pairs that are computationally efficient enough to optimize the global problem each time a new submap is added. We deploy the proposed system on a hexacopter Micro Aerial Vehicle (MAV) with an Intel i7-8650U CPU in two realistic scenarios: mapping a large-scale area using a 3D LiDAR, and mapping an industrial space using an RGB-D camera. In the large-scale outdoor experiments, the system optimizes a 120x80 m map in less than 4 s and produces absolute trajectory RMSEs of less than 1 m over 400 m trajectories. Our complete system, called voxgraph, is available as open source 1 .

Towards High-Performance Solid-State-LiDAR-Inertial Odometry and Mapping

“…For large-scale odometry and mapping with limited computational resources, advanced map representations can be employed to improve memory as well as runtime efficiency. Potential options include volumetric mapping using TSDF (Truncated Signed Distance Fields) [3] or mapping with geometric primitives (especially in man-made environment [28]).…”

“…So far, solid-state-LiDAR-based odometry has not been well investigated. In [19], the LiDAR odometry system in [10] was adapted to Livox Mid-40 3 , a solid-state LiDAR with a circular FoV of 38.4 • . Compared with its baseline [10], it employs similar feature-based scan-matching and delivers comparable tracking accuracy with improved runtime via parallelization.…”

Dense Decentralized Multi-robot SLAM based on locally consistent TSDF submaps

“…A more extensive evaluation of the method on dataset with longer trajectories is necessary to assess its performances in face of significant drifts. Furthermore, a perspective would be to adapt the ICP-based matching to the Euclidean SDF representation, as proposed in Voxgraph [32]. The submap registration appears to be the bottleneck of the proposed framework and should be made faster.…”

Kimera: From SLAM to spatial perception with 3D dynamic scene graphs

“…ElasticFusion (Whelan et al, 2015) instead deforms a dense map of surfels and GravityFusion (Puri et al, 2017) builds on top of ElasticFusion by enforcing a consistent gravity direction among all the surfels. Voxgraph (Reijgwart et al, 2020) builds a globally consistent volumetric map by applying graph optimization over a set of submap poses and including odometry and loop-closure constraints. Similarly, DynamicFusion (Newcombe et al, 2015), VolumeDeform (Innmann et al, 2016), and Fusion4D (Dou et al, 2016) use a volumetric representation for fusion and deformation.…”