DeepSemanticHPPC: Hypothesis-based Planning over Uncertain Semantic Point Clouds

Han Yutao
Han Yutao
Lin Hubert
Lin Hubert
Banfi Jacopo
Banfi Jacopo
Campbell Mark
Campbell Mark

ICRA, pp. 4252-4258, 2020.

Cited by: 0|Bibtex|Views17|DOI:https://doi.org/10.1109/ICRA40945.2020.9196828
EI
Other Links: arxiv.org|dblp.uni-trier.de|academic.microsoft.com
Weibo:
In this paper we present DeepSemanticHPPC, a novel framework for planning in unstructured outdoor environments while accounting for uncertain terrain types

Abstract:

Planning in unstructured environments is challenging -- it relies on sensing, perception, scene reconstruction, and reasoning about various uncertainties. We propose DeepSemanticHPPC, a novel uncertainty-aware hypothesis-based planner for unstructured environments. Our algorithmic pipeline consists of: a deep Bayesian neural network whi...More

Code:

Data:

0
Introduction
  • Path planning for complex outdoor environments is challenging due to the unstructured nature of environments that do not fall neatly into discretized space.
  • Recent work models outdoor maps for planning with a point cloud [12], which is more flexible and suitable for unstructured scenes; [12] uses traditional lidar sensing which cannot differentiate between different surface types as broadly as computer vision.
Highlights
  • Path planning for complex outdoor environments is challenging due to the unstructured nature of environments that do not fall neatly into discretized space
  • While these techniques can differentiate between simple terrain types, they do not model the inherent uncertainties and ambiguities in complex scenes which make it difficult to differentiate between terrain types
  • In this paper we present DeepSemanticHPPC (Deep Semantic Hypothesis-based Planner over Point Clouds), a novel algorithmic pipeline for planning over uncertain semantic point clouds, which leverages a Bayesian neural network (BNN) [13], [14] to extract principled estimates of segmentation uncertainty
  • TMRk is a 6D pose attached to the terrain surface, τ k ∈ [0, 1] is the associated static traversability value, wk is a parameter vector specifying a short planar trajectory segment connecting TMRk to the pose in the sequence, and κk is the curvature at the beginning of the trajectory segment. wk specifies a trajectory segment as a cubic curvature polynomial [19] evolving along the planar patch defined by the xy plane of the coordinate frame R attached to TMRk
  • In this paper we present DeepSemanticHPPC, a novel framework for planning in unstructured outdoor environments while accounting for uncertain terrain types
  • Our experiments show DeepSemanticHPPC reduces semantic uncertainty in planned paths and increases the safety of paths planned in environments with unsafe terrains
Results
  • Baseline B2 performs significantly better than B1 because of the inclusion. The authors' DeepSemanticHPPC framework is significantly better than the two baselines.
Conclusion
  • In this paper the authors present DeepSemanticHPPC, a novel framework for planning in unstructured outdoor environments while accounting for uncertain terrain types.
  • The authors' experiments show DeepSemanticHPPC reduces semantic uncertainty in planned paths and increases the safety of paths planned in environments with unsafe terrains.
  • The authors plan to implement DeepSemanticHPPC on a robot for physical experiments.
  • Other interesting directions include exploring the ability to build the point cloud online and incorporating geometric uncertainties
Summary
  • Introduction:

    Path planning for complex outdoor environments is challenging due to the unstructured nature of environments that do not fall neatly into discretized space.
  • Recent work models outdoor maps for planning with a point cloud [12], which is more flexible and suitable for unstructured scenes; [12] uses traditional lidar sensing which cannot differentiate between different surface types as broadly as computer vision.
  • Results:

    Baseline B2 performs significantly better than B1 because of the inclusion. The authors' DeepSemanticHPPC framework is significantly better than the two baselines.
  • Conclusion:

    In this paper the authors present DeepSemanticHPPC, a novel framework for planning in unstructured outdoor environments while accounting for uncertain terrain types.
  • The authors' experiments show DeepSemanticHPPC reduces semantic uncertainty in planned paths and increases the safety of paths planned in environments with unsafe terrains.
  • The authors plan to implement DeepSemanticHPPC on a robot for physical experiments.
  • Other interesting directions include exploring the ability to build the point cloud online and incorporating geometric uncertainties
Funding
  • This work is funded by the ONR under the PERISCOPE MURI Grant N00014-17-1-2699
Reference
  • G. J. Stein, C. Bradley, and N. Roy, “Learning over subgoals for efficient navigation of structured, unknown environments,” in Proc. CORL, 2018, pp. 213–222.
    Google ScholarLocate open access versionFindings
  • A. Ivanov and M. E. Campbell, “Uncertainty constrained robotic exploration: An integrated exploration planner,” IEEE T Contr Syst T, vol. 27, pp. 146–160, 2016.
    Google ScholarFindings
  • S. M. LaValle, Planning Algorithms. New York, NY, USA: Cambridge University Press, 2006.
    Google ScholarFindings
  • K. Walas, “Terrain classification and negotiation with a walking robot,” J Intell Robot Syst, vol. 78, pp. 401–423, 2015.
    Google ScholarLocate open access versionFindings
  • K. A. Mahadhir, S. C. Tan, C. Y. Low, R. Dumitrescu, A. T. M. Amin, and A. Jaffar, “Terrain classification for track-driven agricultural robots,” in Proc. SYSINT, 2014, pp. 775 – 782.
    Google ScholarLocate open access versionFindings
  • R. O. Chavez-Garcia, J. Guzzi, L. M. Gambardella, and A. Giusti, “Learning ground traversability from simulations,” IEEE RAL, vol. 3, pp. 1695–1702, 2017.
    Google ScholarLocate open access versionFindings
  • J. A. Delmerico, A. Giusti, E. Mueggler, L. M. Gambardella, and D. Scaramuzza, “”on-the-spot training” for terrain classification in autonomous air-ground collaborative teams,” in Proc. ISER, 2016.
    Google ScholarLocate open access versionFindings
  • R. Manduchi, A. Castano, A. Talukder, and L. Matthies, “Obstacle detection and terrain classification for autonomous off-road navigation,” Auton Robot, vol. 18, no. 1, pp. 81–102, 2005.
    Google ScholarLocate open access versionFindings
  • P. Filitchkin, “Visual terrain classification for legged robots,” Ph.D. dissertation, University of California, Santa Barbara, 2011.
    Google ScholarFindings
  • M. W. Otte, S. G. Richardson, J. Mulligan, and G. Z. Grudic, “Path planning in image space for autonomous robot navigation in unstructured environments,” J Field Robot, vol. 26, pp. 212–240, 2009.
    Google ScholarLocate open access versionFindings
  • D. Ferguson and A. Stentz, “Using interpolation to improve path planning: The field D* algorithm,” J Field Robot, vol. 23, pp. 79– 101, 2006.
    Google ScholarLocate open access versionFindings
  • P. Krusi, P. Furgale, M. Bosse, and R. Siegwart, “Driving on point clouds: Motion planning, trajectory optimization, and terrain assessment in generic nonplanar environments,” J Field Rob, vol. 34, no. 5, pp. 940–984, 2017.
    Google ScholarLocate open access versionFindings
  • Y. Gal and Z. Ghahramani, “Dropout as a Bayesian approximation: Representing model uncertainty in deep learning,” in Proc. ICML, 2016.
    Google ScholarLocate open access versionFindings
  • A. Kendall and Y. Gal, “What uncertainties do we need in bayesian deep learning for computer vision?” in Proc. NIPS, 2017, pp. 5574– 5584.
    Google ScholarLocate open access versionFindings
  • R. Border, J. D. Gammell, and P. Newman, “Surface edge explorer (SEE): Planning next best views directly from 3d observations,” in Proc. ICRA, 2018, pp. 6116–6123.
    Google ScholarLocate open access versionFindings
  • B. Hepp, D. Dey, S. N. Sinha, A. Kapoor, N. Joshi, and O. Hilliges, “Learn-to-score: Efficient 3d scene exploration by predicting view utility,” in Proc. ECCV, 2018, pp. 437–452.
    Google ScholarLocate open access versionFindings
  • J. Daudelin and M. Campbell, “An adaptable, probabilistic, next best view algorithm for reconstruction of unknown 3d objects,” IEEE RAL, vol. 2, no. 3, pp. 1540–1547, 2017.
    Google ScholarLocate open access versionFindings
  • S. M. LaValle, “Rapidly-exploring random trees: A new tool for path planning,” 1998.
    Google ScholarFindings
  • B. Nagy and A. Kelly, “Trajectory generation for car-like robots using cubic curvature polynomials,” in Proc. FSR, 2001.
    Google ScholarLocate open access versionFindings
  • C. R. Qi, H. Su, K. Mo, and L. J. Guibas, “Pointnet: Deep learning on point sets for 3d classification and segmentation,” in Proc. CVPR, 2017, pp. 77–85.
    Google ScholarLocate open access versionFindings
  • C. R. Qi, L. Yi, H. Su, and L. J. Guibas, “Pointnet++: Deep hierarchical feature learning on point sets in a metric space,” in Proc. NIPS, 2017, pp. 5099–5108.
    Google ScholarLocate open access versionFindings
  • Y. Zhou and O. Tuzel, “Voxelnet: End-to-end learning for point cloud based 3d object detection,” in Proc. CVPR, 2018, pp. 4490–4499.
    Google ScholarLocate open access versionFindings
  • C. R. Qi, W. Liu, C. Wu, H. Su, and L. J. Guibas, “Frustum pointnets for 3d object detection from rgb-d data,” in Proc. CVPR, 2018, pp. 918–927.
    Google ScholarLocate open access versionFindings
  • Y. Wang, Y. Sun, Z. Liu, S. E. Sarma, M. M. Bronstein, and J. M. Solomon, “Dynamic graph cnn for learning on point clouds,” arXiv preprint arXiv:1801.07829, 2018.
    Findings
  • A. X. Chang, T. Funkhouser, L. Guibas, P. Hanrahan, Q. Huang, Z. Li, S. Savarese, M. Savva, S. Song, H. Su, et al., “Shapenet: An information-rich 3d model repository,” arXiv preprint arXiv:1512.03012, 2015.
    Findings
  • T. Hackel, N. Savinov, L. Ladicky, J. Wegner, K. Schindler, and M. Pollefeys, “SEMANTIC3D.NET: A new large-scale point cloud classification benchmark,” ISPRS Annals, vol. 4, no. 1, pp. 91–98, 2016.
    Google ScholarLocate open access versionFindings
  • A. Kirillov, K. He, R. Girshick, C. Rother, and P. Dollar, “Panoptic segmentation,” in Proc. CVPR, 2019, pp. 9404–9413.
    Google ScholarLocate open access versionFindings
  • H. Caesar, J. Uijlings, and V. Ferrari, “Coco-stuff: Thing and stuff classes in context,” in Proc. CVPR, 2018, pp. 1209–1218.
    Google ScholarLocate open access versionFindings
  • B. Zhou, H. Zhao, X. Puig, S. Fidler, A. Barriuso, and A. Torralba, “Scene parsing through ade20k dataset,” in Proc. CVPR, 2017, pp. 633–641.
    Google ScholarLocate open access versionFindings
  • S. Bell, P. Upchurch, N. Snavely, and K. Bala, “Material recognition in the wild with the materials in context database,” in Proc. CVPR, 2015, pp. 3479–3487.
    Google ScholarLocate open access versionFindings
  • L.-C. Chen, Y. Zhu, G. Papandreou, F. Schroff, and H. Adam, “Encoder-decoder with atrous separable convolution for semantic image segmentation,” in Proc. ECCV, 2018, pp. 801–818.
    Google ScholarLocate open access versionFindings
  • M. Ester, H. Kriegel, J. Sander, and X. Xu, “A density-based algorithm for discovering clusters in large spatial databases with noise.” in Proc. KDD, 1996, pp. 226–231.
    Google ScholarLocate open access versionFindings
  • J. Yen, “Finding the k shortest loopless paths in a network,” Manage Sci, vol. 17, no. 11, pp. 712–716, 1971.
    Google ScholarLocate open access versionFindings
  • B. Zhou, A. Lapedriza, A. Khosla, A. Oliva, and A. Torralba, “Places: A 10 million image database for scene recognition,” IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017.
    Google ScholarLocate open access versionFindings
  • F. Chollet, “Xception: Deep learning with depthwise separable convolutions,” in Proc. CVPR, 2017, pp. 1251–1258.
    Google ScholarLocate open access versionFindings
  • S. Shah, D. Dey, C. Lovett, and A. Kapoor, “Airsim: Highfidelity visual and physical simulation for autonomous vehicles,” in Field and Service Robotics, 2017. [Online]. Available: https://arxiv.org/abs/1705.05065
    Findings
  • P. Cignoni, M. Callieri, M. Corsini, M. Dellepiane, F. Ganovelli, and G. Ranzuglia, “Meshlab: an open-source mesh processing tool.” in Eurographics Italian chapter conference, vol. 1, 2008, pp. 129–136.
    Google ScholarFindings
Full Text
Your rating :
0

 

Tags
Comments