Title Page
Contents
Abstract 12
Chapter 1. Introduction 14
Chapter 2. Related Work 20
Chapter 3. Semantic Knowledge Modeling for Multi-robot System 29
3.1. Semantic Knowledge for Advanced Robot System 29
3.2. Semantic Knowledge Modeling for Multi-robot System 32
3.2.1. Relation Knowledge Modeling 32
3.2.2. TOSM-based Environment Element Modeling 36
3.3. Reasoning Rule Modeling for Multi-robot System 41
Chapter 4. Semantic Knowledge-based Hierarchical Task Modeling 46
4.1. Hierarchy Task Modeling with Semantic Knowledge 46
4.1.1. Task Modeling for Hierarchical Planning Structure 46
4.1.2. PDDL Representation of Hierarchy Tasks 49
Chapter 5. Hierarchical Semantic AI Planning for Multi-robot Task Execution 60
5.1. Overview 60
5.2. Hierarchical Planning Approach for Multi-robot 65
5.2.1. Task Planning 65
5.2.2. Task Re-planning 73
5.2.3. TOSM-based Autonomous Navigation 76
Chapter 6. Experiment 89
6.1. Experimental Environments 90
6.2. Experimental Scenarios 91
6.2.1. Semantic Knowledge-based Multi-robot Planning 91
6.2.2. Multi-robot Mission Planning 94
6.3. Experimental Results 98
6.3.1. Semantic Knowledge Test 98
6.3.2. Multi-robot Mission Planning 110
6.4. Applications 128
6.4.1. Warehouse Scenario 129
6.4.2. Campus Scenario 133
6.4.3. ACS Scenario 141
Chapter 7. Conclusion 148
References 152
논문요약 157
Table 1. Comparison for AI planning based robot system. 22
Table 2. Object properties. 34
Table 3. Data properties. 37
Table 4. Example of the task list defined by the hierarchy. 48
Table 5. Representation of Delivery task with PDDL. 51
Table 6. Representation of Guidance task with PDDL. 51
Table 7. Representation of Move task with PDDL. 52
Table 8. Representation of MoveSame task with PDDL. 53
Table 9. Representation of Take task with PDDL. 55
Table 10. Representation of Give task with PDDL. 55
Table 11. Representation of GoToPlace task with PDDL. 57
Table 12. Representation of PickUp task with PDDL. 57
Table 13. Feature matching algorithm pseudo code. 83
Table 14. Connection zone detecting algorithm pseudo code. 87
Table 15. Experimental cases of multi-robot delivery mission. 96
Table 16. Comparison of task planning results with and without knowledge. 108
Table 17. Experimental results of multi-robot task planning. 117
Table 18. Experimental results of multi-robot task re-planning. 124
Table 19. Re-planned behavior sequence with the centralized system. 126
Table 20. Re-planned behavior sequence with the proposed system. 127
Table 21. Planned behavior sequence for logistic scenario. 130
Table 22. Designed specifications of virtual robots. 135
Table 23. Case of designed delivery tasks. 136
Table 24. Representation of task allocation action with PDDL. 143
Table 25. Representation of Charging task with PDDL. 143
Table 26. Representation of Unload coarse-level task with PDDL. 144
Table 27. Representation of Unload fine-level task with PDDL. 144
Figure 1. Triplet Ontological Semantic Model (TOSM) 29
Figure 2. Classification of environments. 31
Figure 3. Comparison of when object properties activate and deactivate during the navigation. (a)When "isLocatedAt" property is activated/deactivated by... 35
Figure 4. Hierarchical relationships of Place in the semantic database. 39
Figure 5. Place & Task Hierarchy Structure. 47
Figure 6. Extending PDDL actions from the prime action "GoTo" 58
Figure 7. Semantic Knowledge-based Multi-robot System Architecture. 60
Figure 8. Semantic Navigation Framework for Multi-robot System 62
Figure 9. Schematic diagram of task planning structure. (a) Centralized Planning Structure. (b) Proposed hybrid structure. 63
Figure 10. Semantic Knowledge-based Hierarchical Task Planning Structure. 65
Figure 11. Example of the multi-robot working environment. 66
Figure 12. Semantic domain-based coarse goal generating step. 67
Figure 13. Coarse-level task-based multi-robot coarse planning step to perform coarse goal lists. 68
Figure 14. Plan restructuring step to group duplicate robots working in the same place at the same time. 70
Figure 15. Fine planning steps to create detailed behavioral sequences for each robot group. 71
Figure 16. Visualization of fine planning results for Group1. 72
Figure 17. Re-planning method in centralized planning approach. 73
Figure 18. Case of re-planning with specific robot group in proposed planning approach. 74
Figure 19. Case of re-planning with all robots in proposed planning approach. 75
Figure 20. System flow diagram of robot. 76
Figure 21. Predicting motion based on robot speed. 79
Figure 22. Differential type robot. 80
Figure 23. Tricycle type robot. 80
Figure 24. Omnidirectional type robot. 81
Figure 25. Simulation environments. (a) Whole driving place. (b) Cross-shaped stem-place "area" . (c) Square-shaped leaf-place "corridor" . (d) Robot... 90
Figure 26. Semantic knowledge test scenario. (a) Crossover situation. (b) Multiple agents charging situation. (c) Different boxes delivery scenario. 92
Figure 27. Example environment case of delivery task. 94
Figure 28. Task execution failure case. (a) Robot path blocking case (b) Robot breakdown case. 97
Figure 29. Planning result without semantic knowledge. (a) Multi-robot task planning result. (b) Task implementing result with multi-robot. 99
Figure 30. Multi-robot planning results with knowledge of "occupancy" . (a) Multi-robot task planning result. (b) Task implementing result with multi-robot. 101
Figure 31. Planned behavior sequences without knowledge of "usage" . (a) Planned sequence with short charging time. (b) Planned sequence with long... 102
Figure 32. Planned behavior sequences with knowledge of "usage" . (a) Planned sequence with short charging time. (b) Planned sequence with long... 104
Figure 33. Planned behavior sequence without task knowledge. 105
Figure 34. Planned behavior sequence with task knowledge. (a) When a specific item is over the robot's payload limit. (b) When all items within the robot's... 106
Figure 35. Operation of the multi-robot system for delivery missions. (a) Execution of delivery missions. (b) Completion of delivery missions. 111
Figure 36. Planned behavior sequence with the centralized structure. 112
Figure 37. Coarse planning result with the proposed structure. 114
Figure 38. Planned task sequences of each group with the proposed structure. (a) Group1 planning result. (b) Group2 planning result. 115
Figure 39. Graph of used place instance quantity for multi-robot task planning (a) Test results with 2 robots. (b) Test results with 3 robots. (c) Test results... 119
Figure 40. Graph of required time for multi-robot task planning. (a) Test results with 2 robots. (b) Test results with 3 robots. (c) Test results with 4 robots. 120
Figure 41. Re-planning in the path blocking case. (a) Path-blocking situation. (b) Driving a detour route. (c) Completing the delivery mission. 121
Figure 42. Re-planning in the robot breakdown case. (a) When a robot breakdown occurred. (b) Re-placing another robot to accomplish the mission.... 122
Figure 43. Logistics scenario environment. (a) Identified areas of the environment (b) Visualized semantic map. 129
Figure 44. The part of planned behavior sequence for logistic scenario. 132
Figure 45. Logistic task execution based on proposed multi-robot system. 132
Figure 46. Test environment with various things and complex places. 134
Figure 47. Semantic map of campus. 135
Figure 48. Semantic TOSM instances in the campus environments. 136
Figure 49. Generated semantic relationship knowledge of each delivery task. (a) Generated "canPickup" knowledge for the Case1. (b) Generated... 137
Figure 50. Delivery task planning results. (a) Planned behavior sequences in the Case1. (b) Planned behavior sequences in the Case2. 138
Figure 51. Semantic knowledge-based multi-robot delivery task execution. (a) Navigating the robot for each task assigned in the Case 1. (b) Pick up a box... 140
Figure 52. ACS scenario environment. (a) Defined places for robots to operate. (b) Grouped areas based on where the robot is working. 142
Figure 53. Planned coarse-level tasks in ACS scenario. 146
Figure 54. Planned fine-level tasks in ACS scenario. 147