Semantic Terrain Segmentation in the Navigation Vision of Planetary Rovers—A Systematic Literature Review

Kuang, Boyu; Gu, Cheng-Zhen; Rana, Zeeshan A.; Zhao, Yifan; Sun, Shuang; Nnabuife, Somtochukwu Godfrey

doi:10.3390/s22218393

Cited by 10 publications

(5 citation statements)

References 96 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…That is, a detailed 3D model of the rover exploring the target region should be simulated, by SPIS software (Version 5.2.4) for example, before the launch of the CE-7 and during the mission operation. Therefore, software capable of robust real-time path planning for the rover [32] can be developed for designed multipoint wide-area and longitudinal sampling modes.…”

Section: Discussionmentioning

confidence: 99%

Numerical Simulation of the Lunar Polar Environment: Implications for Rover Exploration Challenge

et al. 2023

View full text Add to dashboard Cite

The lunar polar regions are key areas for future exploration due to the long-term continuous illumination and persistently shadowed regions that can cold trap abundant water and other volatiles. However, the complex terrain, dynamic lighting, and solar wind-induced electric-field environment present multiple challenges for polar investigation and sampling missions. China’s Chang’E-7 (CE-7) will explore the Moon’s south polar region in 2026. One of the scientific goals is to drill samples in a wide area with a rover for in situ analysis. This study analyzes the engineering constraints of the polar illumination condition, slopes, and electric field for landing and sampling-site selection. Then, we create a 3D model of CE-7’s lunar rover in three operating environments by employing the Spacecraft Plasma Interaction Software, with the rover sampling (i) on a flat surface, (ii) in a shadow, and (iii) near a meter-scale crater under different solar altitude angles. The results show that the rover can be charged to different potentials under the combined effects of solar wind incident angles and surrounding terrains. We find that a favorable traversing and/or sampling site of the rover for future polar exploration is in the upwind direction of a bulge (positively elevated terrains, such as the lander or boulders) or crater, which will cause a minimum charging effect on the rover. Our results have important implications for minimizing the risk of charging effects and guiding the lunar polar region exploration.

show abstract

Section: Discussionmentioning

confidence: 99%

Numerical Simulation of the Lunar Polar Environment: Implications for Rover Exploration Challenge

et al. 2023

View full text Add to dashboard Cite

show abstract

“…Image-based semantic segmentation has been successfully applied to various robotic applications, including terrain classification [29], [30]. To further improve the precision, various studies have explored the fusion of RGB and depth data for multi-modal semantic segmentation [31], [32].…”

Section: Multi-modal Semantic Segmentationmentioning

confidence: 99%

Spatial Group-Wise Enhance: Enhancing Semantic Feature Learning in CNN

Yang

2023

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Mobile ground robots require perceiving and understanding their surrounding support surface to move around autonomously and safely. The support surface is commonly estimated based on exteroceptive depth measurements, e.g., from LiDARs. However, the measured depth fails to align with the true support surface in the presence of high grass or other penetrable vegetation. In this work, we present the Semantic Pointcloud Filter (SPF), a Convolutional Neural Network (CNN) that learns to adjust LiDAR measurements to align with the underlying support surface. The SPF is trained in a semi-selfsupervised manner and takes as an input a LiDAR pointcloud and RGB image. The network predicts a binary segmentation mask that identifies the specific points requiring adjustment, along with estimating their corresponding depth values. To train the segmentation task, 300 distinct images are manually labeled into rigid and non-rigid terrain. The depth estimation task is trained in a self-supervised manner by utilizing the future footholds of the robot to estimate the support surface based on a Gaussian process. Our method can correctly adjust the support surface prior to interacting with the terrain and is extensively tested on the quadruped robot ANYmal. We show the qualitative benefits of SPF in natural environments for elevation mapping and traversability estimation compared to using raw sensor measurements and existing smoothing methods. Quantitative analysis is performed in various natural environments, and an improvement by 48% RMSE is achieved within a meadow terrain.

show abstract

“…Image-based semantic segmentation has been successfully applied to various robotic applications, including terrain classification [25], [26]. To further improve the precision, various studies have explored the fusion of RGB and depth data for multi-modal semantic segmentation [27], [28].…”

Section: Multi-modal Semantic Segmentationmentioning

confidence: 99%

Seeing Through the Grass: Semantic Pointcloud Filter for Support Surface Learning

Li,

Yang,

Frey

et al. 2023

IEEE Robot. Autom. Lett.

View full text Add to dashboard Cite

Mobile ground robots require perceiving and understanding their surrounding support surface to move around autonomously and safely. The support surface is commonly estimated based on exteroceptive depth measurements, e.g., from LiDARs. However, the measured depth fails to align with the true support surface in the presence of high grass or other penetrable vegetation. In this work, we present the semantic pointcloud filter (SPF), a convolutional neural network (CNN) that learns to adjust LiDAR measurements to align with the underlying support surface. The SPF is trained in a semi-self-supervised manner and takes as an input a LiDAR pointcloud and RGB image. The network predicts a binary segmentation mask that identifies the specific points requiring adjustment, along with estimating their corresponding depth values. To train the segmentation task, 464 distinct images are manually labeled into rigid and non-rigid terrain. The depth estimation task is trained in a self-supervised manner by utilizing the future footholds of the robot to estimate the support surface based on a Gaussian process. Our method can correctly adjust the support surface prior to interacting with the terrain and is extensively tested on the quadruped robot ANYmal. We show the qualitative benefits of SPF in natural environments for elevation mapping and traversability estimation compared to using raw sensor measurements and existing smoothing methods. Quantitative analysis is performed in various natural environments, and an improvement by 48% RMSE is achieved within a meadow terrain.

show abstract

Semantic Terrain Segmentation in the Navigation Vision of Planetary Rovers—A Systematic Literature Review

Cited by 10 publications

References 96 publications

Numerical Simulation of the Lunar Polar Environment: Implications for Rover Exploration Challenge

Numerical Simulation of the Lunar Polar Environment: Implications for Rover Exploration Challenge

Spatial Group-Wise Enhance: Enhancing Semantic Feature Learning in CNN

Seeing Through the Grass: Semantic Pointcloud Filter for Support Surface Learning

Contact Info

Product

Resources

About