ROBOT-CENTRIC ELEVATION MAPPING WITH UNCERTAINTY ESTIMATES
By Péter Fankhauser, Michael Bloesch, Christian Gehring, Marco Hutter, Roland Yves Siegwart. 2014
212 citations recorded by the source corpus.
This paper addresses the local terrain mapping process for an autonomous robot. Building upon an onboard range measurement sensor and an existing robot pose estimation, we formulate a novel elevation mapping method from a robot-centric perspective. This formulation can explicitly handle drift of the robot pose estimation which occurs for many autonomous robots. Our mapping approach fully incorporates the distance sensor measurement uncertainties and the six-dimensional pose covariance of the robot. We introduce a computationally efficient formulation of the map fusion process, which allows for mapping a terrain at high update rates. Finally, our approach is demonstrated on a quadrupedal robot walking over obstacles.
Subfields: Aerospace Engineering, Biomedical Engineering, Computer Vision and Pattern Recognition
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Aerospace Engineering
- Jitendra Malik
- Marco Hutter
- Deepak Pathak
- Vladlen Koltun
- Ashish Kumar
- Zipeng Fu
- Jemin Hwangbo
- Ananye Agarwal
In Biomedical Engineering
- Marco Hutter
- Jemin Hwangbo
- Jitendra Malik
- Deepak Pathak
- Ashish Kumar
- Vladlen Koltun
- Lorenz Wellhausen
- Zipeng Fu
In Computer Vision and Pattern Recognition
- Jitendra Malik
- Luc Van Gool
- Vladlen Koltun
- Trevor Darrell
- Ming-Hsuan Yang
- Dacheng Tao
- Xiaogang Wang
- Bernt Schiele
Citations in the corpus
- Learning quadrupedal locomotion over challenging terrain (2020, background): “Existing terrain estimation pipelines that use cameras or LiDAR (26) failed in environments with snow (Fig.”
- Probabilistic Terrain Mapping for Mobile Robots With Uncertain Localization (2018, methodology): “On the other side, the approach in [14] insufficiently captures the multi-modal distribution in the map fusion process, leading to ‘overshooting’ artifacts at edges.”
- Legged Locomotion in Challenging Terrains using Egocentric Vision (2022, methodology): “Solutions to this include incorporating uncertainty in the elevation map [50, 74, 75] and 252 simulating errors at training time to make the walking policy robust to them [9].”
- Robot Parkour Learning (2023, methodology): “Classical modular methods rely on decoupled visual perception and control pipelines, where the elevation maps [75, 76, 77, 78, 79, 80, 81], traversability maps [82, 83, 84, 85], or state estimators [86, 87, 88, 89, 90, 91, 92, 93, 94, 95] are constructed as intermediate representations for downstrea”
- ANYmal - toward legged robots for harsh environments (2017, background): “In static gaits, individual footholds can be selected purely based on the system kinematics, desired walking speed, as well as the estimated ground elevation from the lidar sensors [30].”
- Contact-aided invariant extended Kalman filtering for robot state estimation (2019, background): “This is unacceptable for global mapping and planning algorithms; however, local elevation maps can still be obtained (Fankhauser et al., 2014).”
- Legged Robot State-Estimation Through Combined Forward Kinematic and Preintegrated Contact Factors (2017, background): “Accurate estimates of the robot’s state and environment are essential prerequisites for both stable control and motion planning [4], [5], [6].”