A Caltech Library Service

Neural Scene Representation for Locomotion on Structured Terrain

Hoeller, David and Rudin, Nikita and Choy, Christopher and Anandkumar, Animashree and Hutter, Marco (2022) Neural Scene Representation for Locomotion on Structured Terrain. IEEE Robotics and Automation Letters, 7 (4). pp. 8667-8674. ISSN 2377-3766. doi:10.1109/LRA.2022.3184779.

[img] PDF - Accepted Version
See Usage Policy.


Use this Persistent URL to link to this item:


We propose a learning-based method to reconstruct the local terrain for locomotion with a mobile robot traversing urban environments. Using a stream of depth measurements from the onboard cameras and the robot’s trajectory, the algorithm estimates the topography in the robot’s vicinity. The raw measurements from these cameras are noisy and only provide partial and occluded observations that in many cases do not show the terrain the robot stands on. Therefore, we propose a 3D reconstruction model that faithfully reconstructs the scene, despite the noisy measurements and large amounts of missing data coming from the blind spots of the camera arrangement. The model consists of a 4D fully convolutional network on point clouds that learns the geometric priors to complete the scene from the context and an auto-regressive feedback to leverage spatio-temporal consistency and use evidence from the past. The network can be solely trained with synthetic data, and due to extensive augmentation, it is robust in the real world, as shown in the validation on a quadrupedal robot, ANYmal, traversing challenging settings. We run the pipeline on the robot’s onboard low-power computer using an efficient sparse tensor implementation and show that the proposed method outperforms classical map representations.

Item Type:Article
Related URLs:
URLURL TypeDescription Paper
Hoeller, David0000-0001-8010-9011
Rudin, Nikita0000-0001-5893-0348
Choy, Christopher0000-0002-6566-3193
Anandkumar, Animashree0000-0002-6974-6797
Hutter, Marco0000-0001-9049-534X
Additional Information:© 2022 IEEE. Manuscript received 24 February 2022; accepted 6 June 2022. Date of publication 20 June 2022; date of current version 18 July 2022. This letter was recommended for publication by Associate Editor D. Sadigh and Editor J. Kober upon evaluation of the reviewers’ comments. This work was supported in part by NVIDIA, the Swiss National Science Foundation (SNSF) under Project 188596, in part by the National Centre of Competence in Research Robotics (NCCR Robotics), and in part by the European Union’s Horizon 2020 Research and Innovation Program under Grant Agreement 780883. This work was also conducted as part of ANYmal Research, a community to advance legged robotics.
Funding AgencyGrant Number
Swiss National Science Foundation (SNSF)188596
National Centre of Competence in Research RoboticsUNSPECIFIED
European Research Council (ERC)780883
Subject Keywords:Representation learning, deep learning for visual perception
Issue or Number:4
Record Number:CaltechAUTHORS:20220714-224603901
Persistent URL:
Official Citation:D. Hoeller, N. Rudin, C. Choy, A. Anandkumar and M. Hutter, "Neural Scene Representation for Locomotion on Structured Terrain," in IEEE Robotics and Automation Letters, vol. 7, no. 4, pp. 8667-8674, Oct. 2022, doi: 10.1109/LRA.2022.3184779
Usage Policy:No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:115593
Deposited By: George Porter
Deposited On:15 Jul 2022 23:28
Last Modified:01 Aug 2022 21:49

Repository Staff Only: item control page