ICRA 2018論文全概覽:接收率40%,深度學習/運動與路徑規劃領跑機器人會議
雷鋒網 AI 科技評論按:本周,ICRA 2018 正式在澳大利亞布里斯班舉辦。作為 IEEE 機器人與自動化學會(IEEE Robotics and Automation Society)的重要會議,學術論文自然成為了不可忽視的一個部分。今年在論文上又有哪些最新進展,又有哪些領域成為最受關注的研究方向?雷鋒網 AI 科技評論根據第一天的開幕式環節及官方發布的數據,整理了相應的重要看點。
機器人學是一個涵蓋領域廣闊、應用方向縱深的學科,從 ICRA 2018 可見一斑。ICRA 2018 共收到來自 800 個研究機構的 3681 位作者的投稿,共計 2586 篇,其中 1981 篇 ICRA 論文, 605 篇 RAL 論文,相較去年有 48% 的增長。
從投遞論文數來看,美國仍然是學術研究的高產大國,以 630 篇論文領跑 ICRA 2018;而位於第二梯隊的則是中國(230 篇)、德國(159 篇)、日本(115 篇);第三梯隊的義大利、法國、英國、韓國均在 80 篇左右。
而從國家每百萬人口的論文平均投遞數來看,人口多的國家顯然就不佔優勢了。憑藉人均近 14 篇論文的驕人成績,第一名被新加坡摘得;隨後是瑞士和澳大利亞。
從論文接收率來看,ICRA 相比起其它學術會議來說還是比較高的,平均值為 40.9%,今年的接收率也延續了往年平均數。而值得一提的是,在上海舉辦的 ICRA 2011 以及在香港舉辦的 ICRA 2014 都呈現了較高的接收率,ICRA 2011 達到了驚人的 52%,而 ICRA 2014 則逼近 50%。兩個接收率低於 40% 的 ICRA 會議分別是 ICRA 2013(德國卡爾斯魯厄)以及不到 35% 的 ICRA 2015(瑞典斯德哥爾摩)。
從論文展示數來看,從 2013 年到 2016 年,過去 5 年的 ICRA 平均有 917 篇展示論文,今年則超過了 1000 篇。
本屆 ICRA 總與會人數超過 2800 人,有 1052 位論文作者在會上分享他們最前沿的研究觀點。近 2000 篇主會論文包含 152 個機器人相關的研究方向;其中 35 篇獲獎提名論文佔總論文數約 3%。
據統計,「Deep Learning in Robotics and Automation(機器人與自動化的深度學習)」共有 124 篇相關文章,以 10% 的比重順理成章地成為了最受關注的熱門研究領域。其中,有 7 篇獲獎提名論文可以歸類於這一領域,它們分別是(後方括弧為獲獎類別):
Interactively Picking Real-World Objects with Unconstrained Spoken Language Instructions(HRI)
LabelFusion: A Pipeline for Generating Ground Truth Labels for Real RGBD Data of Cluttered Scenes(Vision)
Optimization Beyond the Convolution: Generalizing Spatial Relations with End-to-End Metric Learning(Vision)
Learning Object Grasping for Soft Robot Hands(Manipulation)
Learning Robotic Assembly from CAD(Automation)
Online Learning of a Memory for Learning Rates(Conference)
Social Attention: Modeling Attention in Human Crowds(Cognitive)
而排名第二熱門的「Motion and Path Planning(運動與路徑規劃)」則有 107 篇論文,有三篇獲獎提名論文可以歸類於這領域,它們分別是:
Learning Robotic Assembly from CAD(Automation)
PRM-RL: Long-range Robotic Navigation Tasks by Combining Reinforcement Learning and Sampling-based Planning(Service)
Where to Look? Predictive Perception with Applications to Planetary Exploration(Service)
(細心的讀者朋友不難發現,Learning Robotic Assembly from CAD 這篇論文同時覆蓋了兩個最為熱門的領域。)
尾隨其後的 Localization 有 81 篇相關論文,但沒有獲獎提名論文。SLAM、Learning and Adaptive Systems、Multi-Robot Systems 均有超過 60 篇相關論文,也分別有四篇獲獎提名論文屬於對應領域。
不過值得一提的是,機器人學具有明顯的長尾效應。在 120 個研究領域中,前十個熱門研究領域的投遞論文超過了一半。以「機器人與自動化的深度學習」和「運動與路徑規劃」為例,相應的研究人員就超過了 800 位。
在所有論文中,有 7 位教授學者在論文投遞數量上遙遙領先,包括:
賓夕法尼亞大學教授 Vijay Kumar,共有 16 篇論文。Vijay Kumar 教授曾經連續兩年作為 CCF-GAIR 大會的演講嘉賓蒞臨深圳,他的精彩演講給我們留下了深刻印象。往期演講可參見:
AIR 053 | 賓夕法尼亞Vijay Kumar:無人機的「5s」趨勢與挑戰
時隔一年 Vijay Kumar 再登 GAIR 講台,講述自動化的技術與社會挑戰
瑞士蘇黎世聯邦理工學院 Roland Siegwart,14 篇論文;
MIT 教授 Daniela Rus,共 13 篇論文。Daniela Rus 也曾在 2016 年來到首屆 CCF-GAIR 2016 做主題報告。往期演講可參考:
AIR 043| MIT機器人實驗室主任:機器人的12項前沿技術趨勢
義大利技術研究所 Nikos Tsagarakis,共有 11 篇論文;
弗賴堡大學教授 Wolfram Burgard,11 篇論文;
UC 伯克利教授 Sergey Levine,共 10 篇論文;
斯坦福教授 Allison M. Okamura,10 篇論文。
以下為論文提名詳細名單:
Best Conference Paper Award:
Task-specific Sensor Planning for Robotic Assembly Tasks;
(&Best Student Paper Award )Design of an Autonomous Racecar: Perception, State Estimation and System Integration;
(&Best Student Paper Award )A Modular Dielectric Elastomer Actuator to Drive Miniature Autonomous Underwater Vehicles;
Online Learning of a Memory for Learning Rates;
Contact Model Fusion for Event-Based Locomotion in Unstructured Terrains;
Power and control autonomy for high speed locomotion with an insect-scale legged microrobot;
Best Paper Award in Medical Robotics:
A lightweight and efficient portable soft exosuit for paretic ankle assistance in walking after stroke;
Mechanical Rubbing of Blood Clots using Helical Robots under Ultrasound Guidance;
Compact Design of a Hydraulic Driving Robot for Intra-operative MRI-guided Bilateral Stereotactic Neurosurgery;
Best Paper Award in Robot Manipulation:
Learning Modes of Within-hand Manipulation;
Learning Object Grasping for Soft Robot Hands;
(& Multi-Robot)Decentralized Adaptive Control for Collaborative Manipulation;
Best Paper Award on Multi-Robot Systems:
Pairwise Consistent Measurement Set Maximization for Robust Multi-robot Map Merging;
Voronoi-Based Coverage Control of Pan/Tilt/Zoom Camera Networks;
Talk Resource-Efficiently to Me: Optimal Communication Planning for Distributed SLAM Front-Ends;
Best Paper Award in Robot Vision:
LabelFusion: A Pipeline for Generating Ground Truth Labels for Real RGBD Data of Cluttered Scenes;
Optimization Beyond the Convolution: Generalizing Spatial Relations with End-to-End Metric Learning;
Online Photometric Calibration of Auto Exposure Video for Realtime Visual Odometry and SLAM;
Best Paper Award in Cognitive Robotics:
Negotiating with a robot: Analysis of Regulatory Focus Behavior;
Social Attention: Modeling Attention in Human Crowds;
(& Service)Temporal Spatial Inverse Semantics for Robots Communicating with Humans;
Best Paper Award in Service Robotics:
Real-time Semantic Segmentation of Crop and Weed for Precision Agriculture Robots Leveraging Background Knowledge in CNNs;
PRM-RL: Long-range Robotic Navigation Tasks by Combining Reinforcement Learning and Sampling-based Planning;
Where to Look? Predictive Perception with Applications to Planetary Exploration;
ICRA Best Paper Award in Automation:
Accurate and Adaptive In situ Fabrication of an Undulated Wall using an On-Board Visual Sensing System;
Learning Robotic Assembly from CAD;
A Novel Automated Construction Scheme for Efficiently Developing Cloud Manufacturing Services;
Best Paper Award on Human-Robot Interaction (HRI) :
Scaling inertial forces to alter weight perception in virtual reality;
Interactively Picking Real-World Objects with Unconstrained Spoken Language Instructions;
Safety Map: A Unified Representation for Biomechanics Impact Data and Robot Instantaneous Dynamic Properties;
Best Paper Award on Unmanned Aerial Vehicles:
Towards a wind-powered UAV for ocean monitoring: performance, control and validation;
Control Inspired Design of a Fixed-Wing Unmanned Aerial-Aquatic Vehicle;
Onboard Vision-based Control of Agile Suspended Payload Maneuvers;
Design, Modeling and Control of Aerial Robot DRAGON: Dual-rotor-embedded-multilink Robot with the Ability of Multi-deGree-of-freedom Aerial TransformatiON.
35 位論文獲獎提名作者將在會議期間進行相應的論文展示,更多精彩內容敬請期待雷鋒網的專題報道。


※迅雷發布星域雲和迅雷鏈開放平台;智付支付被罰超4000萬|AI金融評論日報
※大疆發布新一代熱成像相機禪思XT2,以及全新的開發包Payload SDK
TAG:雷鋒網 |