Design of real-time safety control method with proximity sensing for human-robot collaboration|RISS 상세보기

다국어 초록 (Multilingual Abstract)

Human–robot collaboration (HRC) in dynamic industrial settings demands three properties at once: responsiveness, stability, and safety. Existing approaches often trade agility for safety or rely on ad hoc safeguards that do not guarantee physical stability near humans or during contact. This thesis presents a real-time safety control architecture that unifies avoidance, tracking, proximity handling, and intentional contact in a single loop by combining passivity guarantees with constrained optimization. The stack integrates variable admittance control (VAC), an energy-tank mechanism, quadratic program (QP), proximity sensing, and fuzzy adaptation.

First, we establish a passivity-preserving VAC framework: the energy tank regulates energy injection during online stiffness and damping modulation, maintaining closed-loop stability while allowing rapid compliance changes. Second, we develop a geometry-scored path selection method that is executed through a unified, priority-weighted QP and tracked with VAC, yielding smooth joint commands across links and agile replanning without heuristic arbitration. Third, we add a constraint-based safety layer—distance-proportional speed limiting, repulsive behavior, and contact-aware deceleration—driven by proximity sensing with fuzzy damping adjustment.

Experiments with static and dynamic obstacles, multi-link motion, and intermittent contact show fast, smooth reactions, reduced contacts, maintained separation, and suppressed velocities. During deliberate contact, end-effector speed converges to zero, ensuring bounded energy exchange. User studies rate the behavior as safe and nonthreatening. The results demonstrate that energy-aware admittance control combined with constrained optimization can deliver practical, passivity-grounded safety for real-world HRC without sacrificing responsiveness.

번역하기

국문 초록 (Abstract)

동적 산업 환경에서의 인간–로봇 협업은 반응성, 안정성, 안전성을 동시에 요구한다. 기존 접근은 민첩성을 안전과 맞바꾸거나 임시 안전장치에 의존해 인간 근접 또는 접촉 시 물리적 안정성을 보장하지 못한다. 본 논문은 수동성 보장과 제약 최적화를 결합해 회피, 추종, 근접 처리, 의도적 접촉을 단일 루프에서 통합하는 실시간 안전 제어 아키텍처를 제시한다. 스택은 가변 어드미턴스 제어, 에너지 탱크, 이차 계획법, 근접 센싱, 퍼지 적응을 통합한다.

첫째, 수동성을 유지하는 가변 어드미턴스 제어 프레임워크를 확립했다. 에너지 탱크가 온라인 강성·감쇠 조절 중 에너지 주입을 조절해 폐루프 안정성을 유지하면서 빠른 순응성 변화를 가능하게 한다. 둘째, 기하학 점수 기반 경로 선택 방법을 개발하고 이를 단일 우선순위 가중 이차 계획법으로 실행하며 가변 어드미턴스 제어로 추종해, 링크 전반에 걸쳐 매끄러운 관절 명령과 휴리스틱 중재 없는 민첩한 재계획을 달성했다. 셋째, 퍼지 감쇠 조절이 결합된 근접 센싱을 바탕으로 거리 비례 속도 제한, 반발 거동, 접촉 인지 감속으로 구성된 제약 기반 안전 레이어를 추가했다.

정적·동적 장애물, 다중 링크 동작, 간헐적 접촉 실험에서 빠르고 부드러운 반응, 접촉 감소, 분리 거리 유지, 속도 억제를 보였다. 의도적 접촉 시 말단효과기 속도는 0으로 수렴해 에너지 교환이 유한하게 유지되었다. 사용자 평가는 동작을 안전하고 위협적이지 않다고 판단했다. 결과적으로 에너지 인식 어드미턴스 제어와 제약 최적화를 결합하면 반응성을 희생하지 않고도 실제 인간–로봇 협업에서 수동성에 기반한 실용적 안전을 제공할 수 있음이 입증되었다.

번역하기

동적 산업 환경에서의 인간–로봇 협업은 반응성, 안정성, 안전성을 동시에 요구한다. 기존 접근은 민첩성을 안전과 맞바꾸거나 임시 안전장치에 의존해 인간 근접 또는 접촉 시 물리적 안...

목차 (Table of Contents)

1 Introduction 1
1.1 Motivation 1
1.2 Background 7
1.2.1 Safety Sensor Technology 7
1.2.2 Proximity Sensor 8

1 Introduction 1
1.1 Motivation 1
1.2 Background 7
1.2.1 Safety Sensor Technology 7
1.2.2 Proximity Sensor 8
1.2.3 Dual-Type Proximity Sensor 10
1.2.4 Reactive Collision Avoidance 12
1.2.4.1 Geometric Approach 13
1.2.4.2 Jacobian-Type Approach 15
1.3 Issue of Reactive Collision Avoidance 31
1.3.1 Drawbacks of Geometric Approach 31
1.3.2 Limitation of Previous Jacobian-Type Approach 32
1.4 Objective of Research 35
1.5 Organization of This Thesis 37
2 Adaptive Stiffness Using Variable Admittance Control Based on Energy Tank 39
2.1 Introduction 39
2.2 Adaptive Stiffness Using Variable Admittance Control 46
2.2.1 Variable Admittance Control 46
2.2.2 Trade-Offs in Obstacle Avoidance and Trajectory Tracking with Stiffness Variation 47
2.2.3 Ensuring Stability of Stiffness Variation Using an Energy Tank 49
2.2.3.1 Passivity Analysis of Variable Admittance Control 49
2.2.3.2 Energy Tank for Passivity Preservation 50
2.3 Experiment 56
2.3.1 Experimental Setup 56
2.3.2 Experimental Tasks 56
2.3.3 Experimental Results 58
2.3.3.1 Stability without Energy Tank 58
2.3.3.2 Response under Different Stiffness Levels 58
2.3.3.3 Comparison of Passivity and Energy Tank Methods 59
2.3.4 Discussion 59
2.4 Conclusion 72
3 Fast Obstacle Avoidance Using Real-Time Trajectory Planning and Weight-Prioritized Quadratic Programming 74
3.1 Introduction 74
3.2 Real-Time Obstacle Avoidance 80
3.2.1 Target Position for Obstacle Avoidance 80
3.2.2 Real-Time Trajectory Planning 85
3.2.3 Acceleration Based on Potential Field with VAC 88
3.3 Self-Sensing Compensation 93
3.4 Experiment 97
3.4.1 Experimental Setup 97
3.4.2 Experimental Tasks 97
3.4.3 Experimental Results 98
3.4.3.1 Static Obstacle Avoidance 98
3.4.3.2 Dynamic Obstacle Avoidance 99
3.4.4 Discussion 99
3.5 Conclusion 108
4 Safety-Critical Reactive Motion Using Constrained Variable Admittance Control with Dual-Type Proximity Sensors 110
4.1 Introduction 110
4.2 Admittance Parameter Control for Obstacle Avoidance and Impact Mitigation 115
4.2.1 Variable Stiffness for Distance Maintenance 115
4.2.2 Variable Damping using Fuzzy Logic 116
4.3 Quadratic Programming-Based Control for Safety Control 123
4.3.1 Joint Constraints 124
4.3.2 Cartesian Velocity Constraints 124
4.3.3 Joint Acceleration Constraints 125
4.3.4 Obstacle Avoidance Constraints 126
4.4 Experiment 129
4.4.1 Experimental Setup 129
4.4.2 Experimental Tasks 129
4.4.3 Experimental Results 131
4.4.3.1 Static Obstacle Avoidance 131
4.4.3.2 Dynamic Obstacle Avoidance(Human) 133
4.4.3.3 Dual-Obstacle Avoidance at a Single Link 133
4.4.3.4 Multi-Link Avoidance(Two Humans) 134
4.4.3.5 Human Contact Safety Evaluation 134
4.4.4 Discussion 135
4.5 Conclusion 152
5 Conclusion 155
Bibliography 159
Abstract 174