IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) 2020

Adaptive Kernel Inference for Dense and Sharp Occupancy Grids

Adaptive Kernel Inference for Dense and Sharp Occupancy Grids

by Youngsun Kwon1, Bochang Moon2, and Sung-Eui Yoon1

1Korea Advanced Institute of Science and Technology (KAIST)

2Gwangju Institute of Science and Technology (GIST)


In this paper, we present a new approach, AKIMap, that uses an adaptive kernel inference for dense and sharp occupancy grid representations. Our approach is based on the multivariate kernel estimation, and we propose a simple, two-stage based method that selects an adaptive bandwidth matrix for an efficient and accurate occupancy estimation. To utilize correlations of occupancy observations given sparse and non-uniform distributions of point samples, we propose to use the covariance matrix as an initial bandwidth matrix, and then optimize the bandwidth matrix by adjusting its scale in an efficient, data-driven way for on-the-fly mapping. We demonstrate that the proposed technique estimates occupancy states more accurately than state-of-the-art methods given equal-data or equal-time settings, thanks to our adaptive inference. Furthermore, we show the practical benefits of the proposed work in on-the-fly mapping and observe that our adaptive approach shows the dense as well as sharp occupancy representations in a real environment.

These figures show the experimental results in the structured scene. We visualize the occupied cells classified based on the occupancy threshold (0.5), where the color represents the elevation from 0.0 m to 2.0 m. The grayscale 2-D image located in the left-bottom represents the occupancy probabilities of cells of a L-shape at the 1.0 m height; colors from white to black represent occupancy probability from the free to occupied states. The number in each parenthesis indicates the amount of used sensor measurements for each algorithm.

These figures show the results for on-the-fly mapping test using a mobile robot. The dotted line in the color figure indicates a trajectory of the robot, and the red and blue boxes show enlarged regions. The image associated with the black box represents the visualization at the viewpoint of follower. As shown in these boxes, ours represents the more sharp and dense surfaces of the wall and the ground in the corridor, compared to the prior work.


Paper: PDF(3.7MB)
Source code: Github, ZIP(2.5MB)
Presentation video: Video(48.6MB)

	title={Adaptive Kernel Inference for Dense and Sharp Occupancy Grids},
	author={Kwon, Youngsun and Moon, Bochang and Yoon, Sung-Eui},
	booktitle={2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},