Occupancy Map Inpainting for Online Robot Navigation

Wei, Minghan; Lee, Daewon; Isler, Volkan; Lee, Daniel

doi:10.1109/icra48506.2021.9561790

Cited by 7 publications

(4 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Wei et al [15] present an alternative approach by limiting the region-of-interest for inpainting. In this approach the robot already has a camera at the height of 1.5m, tilted towards the ground and another camera is placed at the height of 2m with tilt.…”

Section: Related Workmentioning

confidence: 99%

“…To overcome the aforementioned issues, we adopt the approach by Wei et al [15] for predicting the occupancy maps obtained by cameras in the same horizontal plane. We use additional cameras to the left and right side of the robotmounted camera, at the same height and looking straight.…”

Section: Related Workmentioning

confidence: 99%

“…The resulting map from Cam Center , referred to as O c , acts as the input to the network. Similar to Wei et al, [15] we construct the ground truth map O * , as the combination of the maps from the three cameras as follows:…”

Section: A Data Collectionmentioning

confidence: 99%

“…We also train a UNet model for prediction tasks, i.e. image-to-image translation without a discriminator as an equivalent approach to [15] for comparison. In the rest of the discussion, we will refer to the pix2pix approach as UNet-GAN and the prediction approach as UNet-pred.…”

Section: B Network Architecture and Training Detailsmentioning

confidence: 99%

See 3 more Smart Citations

Occupancy Map Prediction for Improved Indoor Robot Navigation

Sharma¹,

Chen²,

Shrivastava³

et al. 2022

Preprint

View full text Add to dashboard Cite

In the typical path planning pipeline for a ground robot, we build a map (e.g., an occupancy grid) of the environment as the robot moves around. While navigating indoors, a ground robot's knowledge about the environment may be limited by the occlusions in its surroundings. Therefore, the map will have many as-yet-unknown regions that may need to be avoided by a conservative planner. Instead, if a robot is able to correctly infer what its surroundings and occluded regions look like, the navigation can be further optimized. In this work, we propose an approach using pix2pix and UNet to infer the occupancy grid in unseen areas near the robot as an image-to-image translation task. Our approach simplifies the task of occupancy map prediction for the deep learning network and reduces the amount of data required compared to similar existing methods. We show that the predicted map improves the navigation time in simulations over the existing approaches.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: A Data Collectionmentioning

confidence: 99%

Section: B Network Architecture and Training Detailsmentioning

confidence: 99%

See 2 more Smart Citations