Learning-Aided 3-D Occupancy Mapping With Bayesian Generalized Kernel Inference

Doherty, Kevin; Shan, Tixiao; Wang, Jinkun; Englot, Brendan

doi:10.1109/tro.2019.2912487

Cited by 34 publications

(32 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [30], BKI has been used on a Bernoulli-distributed random event with beta-distributed prior to model collision in safe high-speed navigation problems and could achieve safe behavior in a novel environment with no relevant training data. BKI was first used in the context of mapping problems in [19], [20], to generalize the discrete counting sensor model [23] to continuous occupancy mapping. Following the same idea, we apply BKI in our semantic counting sensor model and generalize it to continuous semantic mapping.…”

Section: Related Workmentioning

confidence: 99%

Bayesian Spatial Kernel Smoothing for Scalable Dense Semantic Mapping

Gan

Zhang

Grizzle

et al. 2020

IEEE Robot. Autom. Lett.

View full text Add to dashboard Cite

This paper develops a Bayesian continuous 3D semantic occupancy map from noisy point cloud measurements. In particular, we generalize the Bayesian kernel inference model for occupancy (binary) map building to semantic (multi-class) maps. The method nicely reverts to the original occupancy mapping framework when only one occupied class exists in obtained measurements. First, using Categorical likelihood and its conjugate prior distribution, we extend the counting sensor model for binary classification to a multi-class classification problem which results in a unified probabilistic model for both occupancy and semantic probabilities. Secondly, by applying a Bayesian spatial kernel inference to the semantic counting sensor model, we relax the independent grid assumption and bring smoothness and continuity to the map inference. These latter properties enable the method to exploit local correlations present in the environment to predict semantic probabilities in regions unobserved by the sensor while increasing the performance. Lastly, computational efficiency and scalability are achieved by leveraging sparse kernels and a test-data octrees data structure. The evaluations using multiple sequences of stereo camera and LiDAR datasets show that the proposed method consistently outperforms the compared baselines. We also present a qualitative evaluation using data collected by a biped robot platform on the University of Michigan -North Campus.

show abstract

Section: Related Workmentioning

confidence: 99%

Bayesian Spatial Kernel Smoothing for Scalable Dense Semantic Mapping

Gan

Zhang

Grizzle

et al. 2020

IEEE Robot. Autom. Lett.

View full text Add to dashboard Cite

show abstract

“…inference complexity where K is the number of blocks. [13,3] introduces Bayesian Kernel Inference into the mapping problem, further reducing the inference complexity to O(M log N ).…”

Section: Semantic Occupancy Mappingmentioning

confidence: 99%

“…However in the occupancy map with Bayesian inference, training and testing steps are separated, which requires the free space to be represented by some geometries. Selected or sampled points along beams are commonly used in the literature [2,18,13,19,20] to represent free space. [2] uses the closest point on each beam to the query point as a free space data point.…”

Section: Free Space Representation In Occupancy Mappingmentioning

confidence: 99%

Real-time Semantic 3D Dense Occupancy Mapping with Efficient Free Space Representations

Zhong¹,

Peng

2021

Preprint

View full text Add to dashboard Cite

A real-time semantic 3D occupancy mapping framework is proposed in this paper. The mapping framework is based on the Bayesian kernel inference strategy from the literature. Two novel free space representations are proposed to efficiently construct training data and improve the mapping speed, which is a major bottleneck for real-world deployments. Our method achieves real-time mapping even on a consumer-grade CPU. Another important benefit is that our method can handle dynamic scenarios, thanks to the coverage completeness of the proposed algorithm. Experiments on real-world point cloud scan datasets are presented.

show abstract

“…To this end, LiDAR is a good choice as a sensor (Doherty et al , 2019; Gallant and Marshall, 2016; Bosse et al , 2012). LiDAR is more stable than cameras because the noise associated with each distance measurement is independent on the distance and the lighting conditions (Deschaud, 2018).…”

Section: Introductionmentioning

confidence: 99%

GLO-SLAM: a slam system optimally combining GPS and LiDAR odometry

Lin

Zhang

2021

View full text Add to dashboard Cite

Purpose Large-scale and precise three-dimensional (3D) map play an important role in autonomous driving and robot positioning. However, it is difficult to get accurate poses for mapping. On one hand, the global positioning system (GPS) data are not always reliable owing to multipath effect and poor satellite visibility in many urban environments. In another hand, the LiDAR-based odometry has accumulative errors. This paper aims to propose a novel simultaneous localization and mapping (SLAM) system to obtain large-scale and precise 3D map. Design/methodology/approach The proposed SLAM system optimally integrates the GPS data and a LiDAR odometry. In this system, two core algorithms are developed. To effectively verify reliability of the GPS data, VGL (the abbreviation of Verify GPS data with LiDAR data) algorithm is proposed and the points from LiDAR are used by the algorithm. To obtain accurate poses in GPS-denied areas, this paper proposes EG-LOAM algorithm, a LiDAR odometry with local optimization strategy to eliminate the accumulative errors by means of reliable GPS data. Findings On the KITTI data set and the customized outdoor data set, the system is able to generate high-precision 3D map in both GPS-denied areas and areas covered by GPS. Meanwhile, the VGL algorithm is proved to be able to verify reliability of the GPS data with confidence and the EG-LOAM outperform the state-of-the-art baselines. Originality/value A novel SLAM system is proposed to obtain large-scale and precise 3D map. To improve the robustness of the system, the VGL algorithm and the EG-LOAM are designed. The whole system as well as the two algorithms have a satisfactory performance in experiments.

show abstract

Learning-Aided 3-D Occupancy Mapping With Bayesian Generalized Kernel Inference

Cited by 34 publications

References 24 publications

Bayesian Spatial Kernel Smoothing for Scalable Dense Semantic Mapping

Bayesian Spatial Kernel Smoothing for Scalable Dense Semantic Mapping

Real-time Semantic 3D Dense Occupancy Mapping with Efficient Free Space Representations

GLO-SLAM: a slam system optimally combining GPS and LiDAR odometry

Contact Info

Product

Resources

About