Research on obstacle avoidance of DWA algorithm based on kalman filter prediction in dynamic environment

국문 초록 (Abstract)

16 개 동적 장애물 환경 실험에서 제안 방법은 기준 DWA 대비 다음의 성능을 달성했다:
경로 24.5% 단축(26.1m→19.7m), 도착시간 20.0% 감소(55.4 초→44.3 초), 안전거리 167.9%
증가(0.28m→0.75m), 평활도 60.1% 개선(11.8→4.7). 실시간 성능은 8.2ms 계획시간으로
유지되었다. 특히 Method_KF 는 효율성, 안전성, 평활도를 동시에 개선한 유일한 방법이며,
다른 개선 기법들(Method_1, 2, 3)은 경로 길이가 오히려 8.8%-18.2% 증가하고 도착시간도 3.8%-
5.3% 지연되었다. 장애물 속도를 1 배에서 6 배까지 변화시키는 강건성 실험에서 Method_KF 는
100% 작업 완료율과 일관된 안전거리(0.38m-0.75m)를 유지한 반면, Method_3 는 67% 성공률을
보이며 3.0×-4.0× 속도에서 완전 실패하고 6.0× 극한 속도에서 위험한 근접거리(0.12m)를
나타냈다.
주요어: 동적 윈도우 접근법, 칼만 필터, 동적 장애물 회피, 이동 로봇, 내비게이션, 시간
동기화 예측

목차 (Table of Contents)

• I. Introduction 1
• 1.1 Research Background and Significance 1
• 1.2 Dynamic Window Approach and Current Research Status 1
• 1.3 Application of Kalman filtering in Motion Prediction 2
• 1.4 Research Content and Innovations of This Paper 2
• 1.5 Organization of the Thesis 4
• II. Previous Work 5
• 2.1 Introduction to DWA Algorithm 5
• 2.1.1 DWA Algorithm Principle 6
• 2.1.1.1 Velocity Sampling Space 6
• 2.1.1.2 Trajectory Prediction 9
• 2.1.1.3 Trajectory Evaluation 11
• 2.1.2 DWA Algorithm Flow 14
• 2.2 Kalman Filtering 16
• 2.2.1 Basic Introduction to Kalman filtering 16
• 2.2.2 Basic Principles of Calculation 16
• 2.2.3 Integration with the DWA Method 18
• 2.3 Summary of This Chapter 18
• III. Proposed Method: DWA with KF-Based Dynamic Obstacle Prediction and System Design 20
• 3.1 Problem Analysis and Improvement Approach 20
• 3.1.1 Limitations of Traditional DWA in Dynamic Environments 21
• 3.1.1.1 Limitations of the Static Obstacle Assumption 21
• 3.1.1.2 Limitations of Instantaneous Observation-Based Decision-Making 22
• 3.1.1.3 Defects of Dimensional Decoupling in the Evaluation Function 23
• 3.1.2 Overview of Existing Improvement Approaches 24
• 3.1.3 Design Principles and Research Route 25
• 3.2 Design of Comparative Baseline Methods 26
• 3.2.1 Four types of comparison methods 26
• 3.2.2 Common Limitations of Three Baseline Methods 29
• 3.3 Method_KF Proposed in This Paper 29
• 3.3.1 Overall Framework 29
• 3.3.2 Design of KF Bank 30
• 3.4 Time-Aligned Collision Risk Assessment 33
• 3.4.1 Defects of Traditional Methods 33
• 3.4.2 Time-Synchronized Prediction Principle 34
• 3.4.3 Comparison with Traditional Methods. 38
• 3.5 Normalized Evaluation Function Framework 38
• 3.5.1 Reference Value Normalization 38
• 3.5.2 Segmented Safety Distance Evaluation 39
• 3.5.3 Sigmoid Nonlinear Fusion 40
• 3.6 Complete Algorithm Flow 41
• 3.7 Chapter Summary 43
• IV. Experiments and Results Analysis 45
• 4.1 Experimental Objectives and Verification Plan 45
• 4.1.1 Verification Objectives 45
• 4.1.2 Comparison Baseline Methods 46
• 4.1.3 Evaluation Index System 47
• 4.2 Simulation Platform Construction 47
• 4.2.1 Software and Hardware Environment 47
• 4.2.2 Scene Design 48
• 4.2.3 Parameter Configuration 50
• 4.2.4 Early Stop Rules and Safety Protection 52
• 4.3 Experiment One: Basic Performance Comparison of Five Algorithms 53
• 4.3.1 Experiment Setup Description 53
• 4.3.2 Test Data Analysis 53
• 4.3.3 Relative Analysis of Improvement Range 56
• 4.3.4 Comparison Analysis of Typical Movement Paths 59
• 4.4 Experiment Two: Scanning Test Under Different Speed Conditions 63
• 4.4.1 Purpose of the Experiment 63
• 4.4.2 How the Experiment is Set Up 63
• 4.4.3 Results Obtained from the Experiment 64
• 4.5 Analysis and Discussion of Experiment Results 67
• 4.5.1 Core Advantages of Method_KF 67
• 4.5.2 Performance Analysis of Method_3 68
• 4.5.3 Analysis of Calculation Complexity 69
• 4.5.4 Limits and Suitable Scenarios 70
• 4.5.5 Experimental Conclusions 73
• 4.6 Chapter Summary 73
• V. Conclusion and Future Work 74
• 5.1 Research Summary 75
• References 76