T-RO 제목 분석 인사이트

IEEE Transactions on Robotics (2004–2025) 3,331편의 제목·인용·연도·키워드를 전수 분석해 도출한 실증 규칙
Data Source
본 분석의 T-RO 논문 메타데이터(제목·저자·연도·인용수)는 RoboPaper Atlasgisbi-kim/robopaper-atlas에서 제공하는 T-RO 역사 전체 파싱 데이터를 사용했습니다. Abstract 및 OpenAlex topic cluster 데이터는 RoboPaper Atlas가 의존하는 동일 소스인 OpenAlex에서 추가 수집했습니다 (T-RO ISSN 1941-0468, 2026-04 기준).
분석 기반: DBLP + OpenAlex (DOI-deduped) · 2026-04 기준 인용수 · 연령보정 percentile tier · Semantic topic assignment (논문당 ≤3주제)
3,331
T-RO 논문 전수
22
분석 연도 (2004–2025)
9
중앙 단어 수 (tier 불문)
26.9%
Top10 콜론 사용률

§0한 줄 결론

The T-RO Title DNA
8~10단어짜리 명사구 한 덩어리로, 무엇을 만들었고 무엇을 위한 것인지를 한 눈에 읽힌다. 최상위 논문의 약 1/3은 ACRONYM: Descriptive Phrase 포맷을 쓰며, 콜론·대문자 약어 사용은 top10 tier에서 2배 많다.
[Method/Attribute] [System] for [Task]  — 가장 흔한 T-RO 뼈대
ACRONYM: [Method] [System] for [Task]  — 오픈소스·시스템 논문의 왕도
Design [and Control] of [Novel System] [for Application]  — 하드웨어 축

§1T-RO는 어떻게 성장했는가

2021년을 기점으로 연 150편→360편으로 출판량이 2배 이상 확장. 평균 인용은 연령이 낮을수록 당연히 낮아지므로, tier 분석은 연령 보정 cites/year로 수행했다.

연도 논문 수 평균 인용 중앙값 최대 성장
2004110102.254779
2005126125.084673
2006119111.669934
2007123118.980855
2008133144.383902
2009138102.664749
201097114.477729
2011107118.777759
2012123108.377695
201311974.457419
201413389.248839
201512288.059875
201611777.860500
201711484.945910
201812167.439621
201911160.340427
202012051.934310
202114749.130784
202223446.430430
202328333.921280
202427017.312162
20253644.6289

§2형태 — Top10 논문 제목의 시각적 서명

연령보정 인용 기준 상위 10%인 논문 242편과 하위 30%인 논문 724편(established 2004–2022)을 비교한 결과, "구조적 특징"이 가장 큰 차별자였다.

길이 — "9단어 / 80자" 표준

Tiern평균 단어평균 글자
Top 10%2428.878
Upper4848.980
Mid9648.880
Low 30%7248.477

길이는 tier 불문 거의 동일 — 단어 수가 성공을 결정하지 않는다. 중앙값이 전부 9단어. 6단어 미만이나 13단어 초과는 희소/위험 영역.

콜론과 대문자 약어 — 가장 강한 차별자

Top 10%
26.9%
Upper
15.3%
Mid
12.8%
Low
12.4%
↑ 콜론(":") 사용률. Top10이 Low의 2.2배
Top 10%
17.8%
Upper
12.2%
Mid
9.5%
Low
8.1%
↑ 대문자 약어 이름(ORB-SLAM, VINS, TEASER …) 포함률. Top10이 Low의 2.2배

명명 전통 — T-RO의 약어 계보

ORB-SLAMORB-SLAM2ORB-SLAM3SVOVINS-MonoGVINSStructVIOCanny-VOFrameSLAMCubeSLAMKimera-MultiAirSLAMAQUA-SLAMLCDNetTEASERFASTERRAPTORScan Context++TossingBotPoseRBPFRLOCNEUROExosFreMEnOmni-SwarmAAGEAsynEIOAiDTC-FOREST
접미사 패턴: -SLAM / -VO / -VINS / -Net / -Bot / -Ex / -Sim 가 도메인 신호

§3좋은 단어 vs 나쁜 단어

Top10 tier와 Low tier 제목의 단어 빈도를 log-odds로 비교해, 각 tier에 진짜로 편중된 단어만 추렸다 (min 4 mentions).

✔ Top10에 편중 — 쓸수록 유리한 단어

하드웨어 신체성
continuumsoftpoweredgrippercable-drivenseries
항공·민첩
quadrotorflightlandingaerialdrone
웨어러블·바이오의료
exoskeletonprosthesisrehabilitationlimbinvasivesurgery
현대 방법론
reinforcementdeepneuralpredictivedata-drivenimpedancemonocular
인식·멀티모달
perceptionmultimodaltactilepoint cloudodometryslammappingfusion
시스템·스토리
surveyavoidanceassistanceswarmssafetyphysicaltightlyagiledexterousbiologically

✘ Low에 편중 — 쓸수록 위험한 단어

과도하게 좁은 기하·메커니즘
planarbipednonholonomicGough-Stewartpolyhedralconvexrigidsingularitydecompositionassemblymechanisms
막연한 형용사·수식어
newnovel (단독)generalgeneralizedunifiedimprovedimprovingperformanceadvancedlimited
행정·에디토리얼 (0 인용 보장)
editorialcomments onauthor's replycorrectionsfarewell
오래된 서브토픽
servoinglocalization (classical)nonholonomicflat system
과도한 일반화
task/tasksdatabetweenovermodemultipleselectionstability (단독)

§4첫 단어 — 의외의 신호

"제목의 첫 단어가 무엇인가"는 tier를 강하게 가른다. 특히 TheRobust는 각각 low tier의 1위·2위 시작어.

Top10 tier 첫 단어 TOP 15

Design
9
The
5
Learning
4
Modeling
3
Fast
3
Novel
3
Geometrically
3
Adaptive
3
Survey
2
Data-Driven
2
Variable
2

Low tier 첫 단어 TOP 15

The
20
Robust
10
Modeling
10
Active
10
Design
9
Control
9
Dynamic
8
Adaptive
8
Distributed
7
Guest (editorial)
6
Toward
5

⚠ 실용 규칙

  • The ...로 시작하면 바로 tier가 한 단계 내려간다 — 웬만하면 쓰지 않는다.
  • Robust X 단독 시작은 low의 대표 시그니처 — 뒤에 반드시 참신한 시스템/문제가 붙어야 한다.
  • Design으로 시작하면 하드웨어 논문 신호, top tier에 많다 — 하드웨어면 나쁘지 않은 선택.
  • Toward/Towards는 tier와 상관 약함. "안전 카드"라기보단 "약간 뻥"에 가까움.

§5연도별 주제 트렌드 — 시맨틱 히트맵 (Title + Abstract, OpenAlex topics, 논문당 ≤3주제)

OpenAlex에서 4,000편 T-RO 논문의 abstract까지 수집(84%)하고, OpenAlex의 pre-clustered topic + 확장 keyword rule로 semantic 주제 분류를 수행. 한 논문이 여러 주제에 속하면 최대 3개까지 카운트. 색이 진할수록 해당 연도 논문 중 비중이 높다. 각 행 오른쪽의 숫자는 총 논문 수.

빈도(%): 0 <1.5 1.5–3 3–6 6–10 10–18 18%+
주제 \ 연도total04050607080910111213141516171819202122232425
▸ Core Backbone — 늘 높은 상수/큰 줄기
Motion·Path Planning76726212019151720191717181815141715192321212520
Manipulation·Grasping66318151915151214211817111522121714171821201615
Swarm·Multi-Robot69011131315181717191916161411171919152221221619
SLAM·Localization (generic)581515111420121614131615918101213131714201714
Kinematics·Dynamics·Design490302322141020171414131414149991288967
Legged·Locomotion434381313141091012914915121281311111199
▸ Hardware·Body — 신체성 축의 성장
Soft Robotics6188111410101311161413151818171914161722221614
Continuum Robots99..11212212344324333531
Exoskeleton·Exosuit84...1122121.22322136325
Prosthesis·Amputee395255771188101371115131712139141178
Surgical·Medical112.1.2241535446224335321
Micro·Nano Robotics180.223433578782845465444
▸ Platform — 플랫폼별 흐름
Aerial·UAV·Drone229211112423436569869910128
Underwater·Marine111.124321415731431134333
Autonomous Driving·Ground62.112.11111311111123233
Humanoid58.1.4313311314.2.210111
▸ Sensing·Perception — 감각의 확장
Tactile·Touch Sensing128.23252312...2.325433284
Haptics·Teleoperation2991191011109986858831257455106
Perception·Vision2951186514613658868496975787
Visual-Inertial SLAM·VIO45....1..1.1..1.12114233
LiDAR SLAM·Perception82...111.....12.13233558
Event Camera15.....1..11....1..10.21
Calibration·Identification272...111111.1112.1101.1
Visual Servoing·Vision Control615342224414412.211...1.
▸ Method·AI — 방법론 혁신
Model Predictive Control85.11....1.211.121224557
Impedance·Force Control985111132313.41523213353
Safety·CBF48........1.1.1111212432
Reinforcement Learning168..121..11113222486611912
Learning (general)9221.111.221121111433476
Imitation·LfD36...........1..11201125
Sim-to-Real24...............1.11132
Human-Robot Interaction98..2531.131244233311226
▸ Emerging Frontier — 아직 희소한 새물결
Transformer17..1.......1.......0022
Diffusion Policy6.....................2
LLM·Foundation·VLA122......1..1....11...11

이 히트맵이 말해주는 것 — 한 페이지 요약

↑ 뚜렷한 상승 (2019~2025)

  • Reinforcement Learning: 1% → 12% (12배)
  • Model Predictive Control: 1% → 7%
  • LiDAR SLAM: 0% → 8%
  • Aerial/UAV: 3% → 10% (2015년 이후 꾸준)
  • Soft Robotics: 10% → 22% (2022 정점)
  • Imitation·LfD: 0% → 5% (2025 점프)
  • Exoskeleton·Exosuit: 2018부터 안정적 3~6%
  • Safety·CBF: 2020 이후 1~4% 안착

↓ 뚜렷한 하락 / 희석

  • Kinematics·Dynamics·Design: 30% (2004) → 7% (2025)
  • Visual Servoing: 5% → 0%
  • Humanoid: 3~4% → 1% (legged에 흡수)
  • Haptics·Teleoperation: 11% → 6% (여전히 존재하지만 비중 하락)
  • Micro·Nano Robotics: 2013~2016 스파크 후 4% 수렴

↔ 안정적 상수 (T-RO의 척추)

  • Motion·Path Planning: 14~26% — 언제나 최상위
  • Manipulation·Grasping: 11~22%
  • Swarm·Multi-Robot: 11~22% — 최근 5년 강화
  • SLAM·Localization: 9~20%
  • Legged·Locomotion: 8~15%
  • Prosthesis·Amputee: 5~17% (2016~2018 피크)

✧ 선점 기회 (Emerging Frontier)

  • Diffusion Policy: 2025에 처음 2%로 등장
  • Transformer: 2024~2025 2%씩만 — 초기 단계
  • LLM·Foundation·VLA: 여전히 <1% — T-RO 제목에는 거의 없음
  • Event Camera: 지속적 희소(1~2%), 핵심 그룹은 있음
  • Sim-to-Real: 1~3%로 자리 잡는 중
  • → 이 5개 중 하나를 T-2/T-5 템플릿과 결합하면 2026 상위 tier 노려볼 수 있음

분석 방법 — hard matching이 아닙니다

이전 버전 (v1, 문제 있던 방법)
  • 제목 단어 하드 매칭 (정확히 문자열 "slam" 포함?)
  • 제목만 봄 (abstract 무시)
  • 한 단어 당 1번 카운트
  • "Koopman Operator for Soft Robots" → "soft" 1개 주제로만 인식
현재 버전 (v2, semantic)
  • OpenAlex에서 전체 abstract 수집 (3,353/4,000 = 83.8% 확보)
  • OpenAlex가 자동 부여한 topic 클러스터와 내 keyword rule을 결합
  • Title + Abstract 양쪽에서 synonym·확장 키워드까지 scan
  • 한 논문에 최대 3개 주제 (구체성 우선순위로 상위 3개)
  • 위 예시는 Soft Robotics + Data-Driven + Continuum 3개로 인식됨

§6복제 가능한 구조 템플릿 (T-1 ~ T-8)

실제 고인용 제목에서 공통 구조를 역설계한 8개 템플릿.

T-1 · 최범용
[Attribute] [System] for [Task/Setting]
Variable Impedance Control of Redundant Manipulators for Intuitive Human-Robot Physical Interaction
Model Predictive Contouring Control for Time-Optimal Quadrotor Flight
Vector Field Path Following for Miniature Air Vehicles
Geometrically Constrained Trajectory Optimization for Multicopters
T-2 · 시스템 논문의 왕도 ★
ACRONYM: [Method] [System] for [Task]
SVO: Semidirect Visual Odometry for Monocular and Multicamera Systems
TEASER: Fast and Certifiable Point Cloud Registration
FASTER: Fast and Safe Trajectory Planner for Navigation in Unknown Environments
RAPTOR: Robust and Perception-Aware Trajectory Replanning for Quadrotor Fast Flight
Kimera-Multi: Robust, Distributed, Dense Metric-Semantic SLAM for Multi-Robot Systems
TossingBot: Learning to Throw Arbitrary Objects With Residual Physics
T-3 · 하드웨어/메커니즘
Design [and Control] of [Novel System] [for Application]
Design and Control of Concentric-Tube Robots
Modeling of Soft Fiber-Reinforced Bending Actuators
A Positive Pressure Universal Gripper Based on the Jamming of Granular Material
T-4 · 메타 논문 (고인용 확률 ↑)
Past, Present, and Future of X / A Survey on X / X: A Review
Past, Present, and Future of Simultaneous Localization and Mapping: Toward the Robust-Perception Age
A Survey on Aerial Swarm Robotics
Robot Collisions: A Survey on Detection, Isolation, and Identification
Data-Driven Grasp Synthesis — A Survey
⚠ 편집부 초청 혹은 공인 선임 저자에게만 통하는 카드. 아무나 흉내내면 역효과.
T-5 · 현대 필수
Learning [X] for [Task] / [X] Learning for [Task]
Learning Stable Nonlinear Dynamical Systems With Gaussian Mixture Models
Model-Based Reinforcement Learning for Closed-Loop Dynamic Control of Soft Robotic Manipulators
Deep Drone Racing: From Simulation to Reality With Domain Randomization
T-6 · Perception / SLAM
[X]-Inertial [System] / Visual-[Y] [System] / [Sensor]-[Sensor] Fusion
VINS-Mono: A Robust and Versatile Monocular Visual-Inertial State Estimator
GVINS: Tightly Coupled GNSS-Visual-Inertial Fusion for Smooth and Consistent State Estimation
Omni-Swarm: A Decentralized Omnidirectional Visual-Inertial-UWB State Estimation System for Aerial Swarms
T-7 · 바이오-모방 로코모션
[Biological Modifier] Robot [for/With] [Gait/Locomotion]
Inchworm Inspired Multimodal Soft Robots With Crawling, Climbing, and Transitioning Locomotion
A Locust-Inspired Robot Capable of Continuous Crawl-Jump-Gliding Locomotion With Optimized Transitional Control
The First Takeoff of a Biologically Inspired At-Scale Robotic Insect
T-8 · 재활/외골격
[Capability] Control of [Device] for [Patient/User Context]
Continuous-Phase Control of a Powered Knee-Ankle Prosthesis: Amputee Experiments Across Speeds and Inclines
Human-Robot Interaction Control of Rehabilitation Robots With Series Elastic Actuators
Human-in-the-Loop Control of Soft Exosuits Using Impedance Learning on Different Terrains

§7반드시 피해야 할 안티패턴

#안티패턴왜 나쁜가실제 Low tier 예
1The + 일반명사 ... 시작Low tier 1번째 단어 1위"The Flying Brick: A Cautionary Note on Testing Flying Robots..."
2A Novel [Method] for [Classical Narrow Problem]Novel 단독 = 내용 희석 신호"A Novel Quantitative Measure for Separation Between Convex Primitives..."
313단어 초과 + 문장형가독성 폭락"Spatial admittance selection conditions for frictionless force-guided assembly of polyhedral parts in single principal contact"
4문장 끝에 주석식 대괄호Errata/correction 신호"Corrections to ... [Jun 07 443-458]"
5맥락 없는 단일 응용어 (yo-yo, coin snap)이국적이지만 검색성 낮음"Return Maps, Parameterization, and Cycle-Wise Planning of Yo-Yo Playing"
6약어만 있고 설명 없음읽는 사람 맥락 상실"DTAR — A Dynamic, Tube-Ascending Robot"
7generalized / unified / improved + classical term공허함 신호"An analytical expression for the generalized forces in multibody Lagrange equations"
8LaTeX 수식 ($…$) 삽입검색·인용 불리"A Novel Iterative Solution to the Perspective-$n$-Point Problem..."
9Comments on, Author's reply, Guest Editorial, Farewell본문 아님 — 정의상 0 인용
10끝에 — A Pilot Study, [extended] 붙임기술 소개보다 주석처럼 읽힘"A Unilateral Active Knee Exoskeleton... — A Pilot Study"

§8생성기 알고리듬 (의사코드)

이 체크리스트는 LLM 프롬프트에 그대로 집어넣어도 된다.

INPUT:
  topic       : 연구 대상 (e.g., "soft gripper", "VIO", "quadruped MPC")
  novelty     : 핵심 기여 1문장 (e.g., "contact-aware diffusion policy")
  system?     : 시스템 이름 (optional — 약어 만들고 싶으면)
  year        : 2026
  venue_style : "tro"

STEP A — 구조 선택:
  if 오픈소스/시스템/라이브러리 → T-2 (ACRONYM: ...)
  elif 하드웨어/메커니즘/actuator → T-3 (Design of ...)
  elif 학습 기반 → T-5 (Learning X for Y) or T-2 if 시스템 이름 있음
  elif sensor fusion/SLAM → T-6 (Visual-Inertial ..., X-VIO ...)
  elif exoskeleton/prosthesis/surgical → T-8
  elif bio-inspired locomotion → T-7
  else → T-1 (Method + System for Task)

STEP B — 단어 선택 필터:
  MUST include ≥1 시대 키워드 (2026):
      {soft, tactile, learning, model predictive, visual-inertial, multimodal,
       perception, lidar, legged, aerial, swarm, exoskeleton, continuum,
       diffusion, foundation, transformer}
  SHOULD include 구체 명사 2개 이상 (task + sensor + platform)
  AVOID 금지 단어 (§3 Low list, §7 Anti-patterns)
  IF 1번째 단어 in {the, robust(단독), novel(단독), a, an} → 재작성

STEP C — 길이·형식:
  target_words = 8..10   (hard range: 6..13)
  target_chars = 70..95
  IF colon used:
     prefix = 1 word (ACRONYM) OR 3..5 words (메타/네러티브)
     suffix = 4..8 words
  ELSE:
     single noun phrase, no period

STEP D — 약어 작명 (T-2):
  length = 3..8 chars
  all-caps OR TitleCase-suffix  (e.g., ORB-SLAM, VINS-Mono)
  letters derivable from suffix phrase
  domain suffix 추천: -SLAM, -VO, -VINS, -Net, -Bot, -Sim, -Ex, -Pose, -Fusion

STEP E — 검증:
  ☐ platform + method + task 3요소 읽히는가?
  ☐ Google Scholar에서 중복 없이 identify 가능한가?
  ☐ 5년 뒤에도 유효할 domain-stable 단어 1개 있는가?
  ☐ low-tier marker ≤ 1?

OUTPUT: 3개 후보 + 각 후보별 템플릿/시대키워드/약어 근거

§9같은 연구, 3가지 템플릿

연구: "진동 기반 tactile sensor + diffusion policy로 곡면 grasping"

T-1 적용
Vibration-Tactile Diffusion Policies for Robust Curved-Surface Grasping
9 words · 70 chars · 시대키워드 3개(diffusiontactilegrasping)
T-2 적용 (시스템명)
VibraGrasp: A Tactile-Conditioned Diffusion Policy for Curved-Surface Manipulation
9 words · 81 chars · 약어 + 콜론 + 현대키워드
T-5 적용 (Learning)
Learning Curved-Surface Grasping From Vibration-Tactile Feedback With Diffusion Policies
10 words · 89 chars · Learning 시작 + 시대키워드 집약

§10제출 전 15초 체크리스트

  • 단어 수 6~13, 글자 수 60~100?
  • 첫 단어가 The/A/An/Robust/Novel 단독이 아님?
  • 제목 안에 platform 1개 + method 1개 + task 1개가 드러남?
  • 약어를 썼다면 콜론 뒤에 풀이 문구 있음?
  • 대시 복합어 1~3개 포함? (너무 많으면 오히려 버거움)
  • §3의 금기어(generalized, planar, assembly …) 1개 이하?
  • §5의 시대 키워드 1~2개 포함 (단 overfit 금지)?
  • 10년 뒤에도 유효할 domain-stable 단어 1개 확보?
  • LaTeX 수식 없음? ($n$, $\alpha$ 같은 것 금지)
  • 끝에 — A Pilot Study, [extended] 같은 접미 태그 없음?

§11레퍼런스 20선 — 연령보정 cpy 상위

이 20개를 소리 내서 3번 읽으면, T-RO 제목의 리듬이 몸에 남는다.

7842021 · TEASER: Fast and Certifiable Point Cloud Registration
4302022 · Past, Present, and Future of Aerial Robotic Manipulators
9102017 · SVO: Semidirect Visual Odometry for Monocular and Multicamera Systems
3412022 · Geometrically Constrained Trajectory Optimization for Multicopters
3362022 · GVINS: Tightly Coupled GNSS-Visual-Inertial Fusion for Smooth and Consistent State Estimation
7472017 · Robot Collisions: A Survey on Detection, Isolation, and Identification
7362017 · Safety Barrier Certificates for Collisions-Free Multirobot Systems
8752015 · Modeling of Soft Fiber-Reinforced Bending Actuators
6212018 · A Survey on Aerial Swarm Robotics
8392014 · Data-Driven Grasp Synthesis — A Survey
8202014 · 3-D Mapping With an RGB-D Camera
4272019 · CubeSLAM: Monocular 3-D Object SLAM
2322022 · Kimera-Multi: Robust, Distributed, Dense Metric-Semantic SLAM for Multi-Robot Systems
2792021 · Data-Driven Control of Soft Robots Using Koopman Operator Theory
3882019 · Model-Based Reinforcement Learning for Closed-Loop Dynamic Control of Soft Robotic Manipulators
2652021 · RAPTOR: Robust and Perception-Aware Trajectory Replanning for Quadrotor Fast Flight
6952012 · A Positive Pressure Universal Gripper Based on the Jamming of Granular Material
4482017 · Proprioceptive Actuator Design in the MIT Cheetah: Impact Mitigation and High-Bandwidth Physical Interaction for Dynamic Legged Robots
2812020 · TossingBot: Learning to Throw Arbitrary Objects With Residual Physics
7492009 · Temporal-Logic-Based Reactive Mission and Motion Planning