The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. ex. Some numerals are expressed as "XNUMX".
Copyrights notice
The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. Copyrights notice
Baru-baru ini, kaedah pembelajaran tetulang mendalam (DRL) telah meningkatkan prestasi tugas navigasi dalaman dipacu sasaran dengan ketara. Walau bagaimanapun, maklumat semantik persekitaran yang kaya masih belum dieksploitasi sepenuhnya dalam pendekatan sebelumnya. Di samping itu, kaedah sedia ada biasanya cenderung terlalu sesuai pada adegan latihan atau objek dalam tugas navigasi dipacu sasaran, menjadikannya sukar untuk digeneralisasikan kepada persekitaran yang tidak kelihatan. Manusia boleh dengan mudah menyesuaikan diri dengan adegan baharu kerana mereka boleh mengenali objek yang mereka lihat dan menaakul kemungkinan lokasi objek sasaran menggunakan pengalaman mereka. Diilhamkan oleh ini, kami mencadangkan model navigasi dipacu sasaran berasaskan DRL, yang dinamakan MVC-PK, menggunakan maklumat Konteks Berbilang Paparan dan Pengetahuan semantik Terdahulu. Ia hanya bergantung pada label semantik objek sasaran dan membolehkan robot mencari sasaran tanpa menggunakan sebarang peta geometri. Untuk melihat maklumat kontekstual semantik dalam persekitaran, pengesan objek digunakan untuk mengesan objek yang terdapat dalam pemerhatian berbilang paparan. Untuk membolehkan keupayaan penaakulan semantik robot mudah alih dalaman, Rangkaian Konvolusi Graf juga digunakan untuk menggabungkan pengetahuan sedia ada. Model MVC-PK yang dicadangkan dinilai dalam persekitaran simulasi AI2-THOR. Keputusan menunjukkan bahawa MVC-PK (1) meningkatkan dengan ketara kebolehan generalisasi silang adegan dan sasaran, dan (2) mencapai prestasi terkini dengan peningkatan 15.2% dan 11.0% dalam Kadar Kejayaan (SR) dan Kejayaan ditimbang mengikut Panjang Laluan (SPL), masing-masing.
Jianbing WU
Peking University
Weibo HUANG
Peking University
Guoliang HUA
Peking University
Wanruo ZHANG
Peking University
Risheng KANG
Department of Mechanical Engineering
Hong LIU
Peking University
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Salinan
Jianbing WU, Weibo HUANG, Guoliang HUA, Wanruo ZHANG, Risheng KANG, Hong LIU, "Semantic Path Planning for Indoor Navigation Tasks Using Multi-View Context and Prior Knowledge" in IEICE TRANSACTIONS on Information,
vol. E106-D, no. 5, pp. 756-764, May 2023, doi: 10.1587/transinf.2022DLP0033.
Abstract: Recently, deep reinforcement learning (DRL) methods have significantly improved the performance of target-driven indoor navigation tasks. However, the rich semantic information of environments is still not fully exploited in previous approaches. In addition, existing methods usually tend to overfit on training scenes or objects in target-driven navigation tasks, making it hard to generalize to unseen environments. Human beings can easily adapt to new scenes as they can recognize the objects they see and reason the possible locations of target objects using their experience. Inspired by this, we propose a DRL-based target-driven navigation model, termed MVC-PK, using Multi-View Context information and Prior semantic Knowledge. It relies only on the semantic label of target objects and allows the robot to find the target without using any geometry map. To perceive the semantic contextual information in the environment, object detectors are leveraged to detect the objects present in the multi-view observations. To enable the semantic reasoning ability of indoor mobile robots, a Graph Convolutional Network is also employed to incorporate prior knowledge. The proposed MVC-PK model is evaluated in the AI2-THOR simulation environment. The results show that MVC-PK (1) significantly improves the cross-scene and cross-target generalization ability, and (2) achieves state-of-the-art performance with 15.2% and 11.0% increase in Success Rate (SR) and Success weighted by Path Length (SPL), respectively.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.2022DLP0033/_p
Salinan
@ARTICLE{e106-d_5_756,
author={Jianbing WU, Weibo HUANG, Guoliang HUA, Wanruo ZHANG, Risheng KANG, Hong LIU, },
journal={IEICE TRANSACTIONS on Information},
title={Semantic Path Planning for Indoor Navigation Tasks Using Multi-View Context and Prior Knowledge},
year={2023},
volume={E106-D},
number={5},
pages={756-764},
abstract={Recently, deep reinforcement learning (DRL) methods have significantly improved the performance of target-driven indoor navigation tasks. However, the rich semantic information of environments is still not fully exploited in previous approaches. In addition, existing methods usually tend to overfit on training scenes or objects in target-driven navigation tasks, making it hard to generalize to unseen environments. Human beings can easily adapt to new scenes as they can recognize the objects they see and reason the possible locations of target objects using their experience. Inspired by this, we propose a DRL-based target-driven navigation model, termed MVC-PK, using Multi-View Context information and Prior semantic Knowledge. It relies only on the semantic label of target objects and allows the robot to find the target without using any geometry map. To perceive the semantic contextual information in the environment, object detectors are leveraged to detect the objects present in the multi-view observations. To enable the semantic reasoning ability of indoor mobile robots, a Graph Convolutional Network is also employed to incorporate prior knowledge. The proposed MVC-PK model is evaluated in the AI2-THOR simulation environment. The results show that MVC-PK (1) significantly improves the cross-scene and cross-target generalization ability, and (2) achieves state-of-the-art performance with 15.2% and 11.0% increase in Success Rate (SR) and Success weighted by Path Length (SPL), respectively.},
keywords={},
doi={10.1587/transinf.2022DLP0033},
ISSN={1745-1361},
month={May},}
Salinan
TY - JOUR
TI - Semantic Path Planning for Indoor Navigation Tasks Using Multi-View Context and Prior Knowledge
T2 - IEICE TRANSACTIONS on Information
SP - 756
EP - 764
AU - Jianbing WU
AU - Weibo HUANG
AU - Guoliang HUA
AU - Wanruo ZHANG
AU - Risheng KANG
AU - Hong LIU
PY - 2023
DO - 10.1587/transinf.2022DLP0033
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E106-D
IS - 5
JA - IEICE TRANSACTIONS on Information
Y1 - May 2023
AB - Recently, deep reinforcement learning (DRL) methods have significantly improved the performance of target-driven indoor navigation tasks. However, the rich semantic information of environments is still not fully exploited in previous approaches. In addition, existing methods usually tend to overfit on training scenes or objects in target-driven navigation tasks, making it hard to generalize to unseen environments. Human beings can easily adapt to new scenes as they can recognize the objects they see and reason the possible locations of target objects using their experience. Inspired by this, we propose a DRL-based target-driven navigation model, termed MVC-PK, using Multi-View Context information and Prior semantic Knowledge. It relies only on the semantic label of target objects and allows the robot to find the target without using any geometry map. To perceive the semantic contextual information in the environment, object detectors are leveraged to detect the objects present in the multi-view observations. To enable the semantic reasoning ability of indoor mobile robots, a Graph Convolutional Network is also employed to incorporate prior knowledge. The proposed MVC-PK model is evaluated in the AI2-THOR simulation environment. The results show that MVC-PK (1) significantly improves the cross-scene and cross-target generalization ability, and (2) achieves state-of-the-art performance with 15.2% and 11.0% increase in Success Rate (SR) and Success weighted by Path Length (SPL), respectively.
ER -