Integrating Deep Reinforcement and Supervised Learning to Expedite Indoor Mapping

Zwecher, Elchanan; Iceland, Eran; R., Levy, Sean; Hayoun, Shmuel Y.; Oren, Gal; Barel, Ariel

doi:10.1109/icra46639.2022.9811861

Cited by 7 publications

(3 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Some sampling-based works predict information gain directly by a regression formulation, with a deep network [15] or a Gaussian process [14]. In [8], a reinforcement learning-based approach is proposed paired with a convolutional network for occupancy grid prediction; [9] extends the approach to a real-world exploration system in the form of a micro-aerial vehicle.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

FloorGenT: Generative Vector Graphic Model of Floor Plans for Robotics

Ericson

2022

2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

View full text Add to dashboard Cite

In this paper, we tackle the challenge of predicting the unseen walls of a partially observed environment as a set of 2D line segments, conditioned on occupancy grids integrated along the trajectory of a 360 • LIDAR sensor. A dataset of such occupancy grids and their corresponding target wall segments is collected by navigating a virtual robot between a set of randomly sampled waypoints in a collection of office-scale floor plans from a university campus. The line segment prediction task is formulated as an autoregressive sequence prediction task, and an attentionbased deep network is trained on the dataset. The sequencebased autoregressive formulation is evaluated through predicted information gain, as in frontier-based autonomous exploration, demonstrating significant improvements over both non-predictive estimation and convolution-based image prediction found in the literature. Ablations on key components are evaluated, as well as sensor range and the occupancy grid's metric area. Finally, model generality is validated by predicting walls in a novel floor plan reconstructed on-the-fly in a real-world office environment.

show abstract

Section: Related Workmentioning

confidence: 99%

“…The network was not pretrained. Several network architectures were evaluated for image-based prediction [8], [30], [43]- [46]. All architectures reached convergence at the same loss value, suggesting that convergence is caused by the intrinsic difficulty of image-based prediction.…”

Section: A Predicted Information Gainmentioning

confidence: 99%

FloorGenT: Generative Vector Graphic Model of Floor Plans for Robotics

Ericson

2022

2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

View full text Add to dashboard Cite

show abstract

“…The precision of the integrated navigation system in the unlocked state can be applied to indoor mapping. (10)(11)(12)(13) Combining the results of existing research and technological development, we proposed a data acquisition method using the combination of ground and vehicle-mounted laser scanning technology and applied it to the mapping of underground garages by summarizing and analyzing the existing methods. The vehicle-mounted laser scanning system based on high-precision inertial navigation quickly obtained a large range of indoor and outdoor data and directly obtained the point cloud in the geodetic coordinate system.…”

Section: Introductionmentioning

confidence: 99%

Space-constrained Mobile Laser-point Cloud Data Acquisition Method

Liu¹,

Yang²,

Zhao³

et al. 2023

Sensors and Materials

View full text Add to dashboard Cite

The mobile laser-point cloud measurement system has been widely used in urban highprecision mapping. In some urban spaces with many high-rise buildings and indoor spaces, high-precision point cloud data cannot be directly obtained due to weak or no global navigation satellite system (GNSS) signals. We propose a space-constrained mobile vehicle-mounted laser scanning data acquisition plan that can effectively utilize the advantages of ground 3D laser scanning and mobile vehicle-mounted laser scanning. In addition, the plan adopts multi-source point cloud data fusion technology to obtain high-precision complete point cloud data both indoors and outdoors. The validity and reliability of the plan are verified by data acquisition experiments, which provide a useful reference for future high-precision mapping.

show abstract