Contents
Statement vii
Abstract xi
Acknowledgements xiii
1 Introduction 1
1.1 Thesis structure and original contributions . . . . 4
1.2 Other contributions . . . . 7
1.3 Publication summary . . . . 8
2 Robot platforms 11 3 Mergeable nervous systems for robots 15 3.1 Enhanced directional self-assembly (EDSA) . . . . 16
3.1.1 The mxRAB device and messages communicated . . . . 17
3.1.2 Recruitment, guidance, and maneuvering algorithms . . . . . 19
3.1.3 Speed, precision and other features . . . . 23
3.1.3.1 Adaptive recruitment . . . . 24
3.1.3.2 Enhanced parallelism . . . . 26
3.1.3.3 Morphology growth in motion . . . . 26
3.2 From EDSA to topology-aware larger morphologies . . . . 26
3.2.1 Controlled morphology formation . . . . 27
3.2.2 Topology and partial failure representation . . . . 28
3.2.3 Single message-based topology update mechanism . . . . 30
3.3 MNS robot control . . . . 30
3.3.1 Scalability . . . . 35
3.3.2 Unprecedented features and self-healing properties . . . . 37
3.3.2.1 Borrowing hardware capabilities of peer robots . . . 37
3.3.2.2 Autonomous adaptation to varying scales and mor- phologies . . . . 38
3.3.2.3 Morphology-independent sensorimotor coordination 38 3.3.2.4 Fault-detection and self-healing properties . . . . . 39
3.4 Related work . . . . 43
3.5 Summary . . . . 45
4 Establishing spatially targeted communication links 47 4.1 Establishing a one-to-one communication link . . . . 48
4.1.1 The iterative elimination process and preliminary trends . . . 48
4.1.2 Markov chain model and model-based analysis . . . . 51 xv
4.1.3 Experiments and results . . . . 55
4.2 Establishing a one-to-many communication link . . . . 58
4.2.1 The iterative growth process . . . . 59
4.2.2 Square lattice distribution-based Markov chain model . . . . 63
4.2.2.1 Deterministic phase model . . . . 63
4.2.2.2 Stochastic phase model . . . . 64
4.2.2.3 Analyzing model predictions and scalability . . . . . 64
4.2.3 Experiments and results . . . . 67
4.3 Discussion . . . . 70
4.4 Related work . . . . 72
4.5 Summary . . . . 73
5 Supervised morphogenesis 75 5.1 Control methodology . . . . 76
5.1.1 Aerial robot . . . . 76
5.1.2 Self-assembling robots . . . . 77
5.2 Case study number 1 . . . . 78
5.2.1 Task and experimental setup . . . . 78
5.2.2 Results . . . . 79
5.3 Case study number 2 . . . . 81
5.3.1 Task and experimental setup . . . . 81
5.3.2 3D environment modeling using heightmaps . . . . 82
5.3.3 Decision-making based on height profiles . . . . 82
5.3.4 Results . . . . 83
5.4 Quantifying performance benefits . . . . 86
5.4.1 Task and experimental setup . . . . 86
5.4.2 Control methodologies . . . . 87
5.4.2.1 Non-cooperative control (NCC) . . . . 88
5.4.2.2 Location-based supervised morphogenesis (LSM) . . 88
5.4.2.3 Supervision based on random group (SRG) . . . . . 89
5.4.3 Experiments and results . . . . 89
5.4.3.1 NCC vs. LSM . . . . 89
5.4.3.2 SRG vs. LSM . . . . 91
5.5 Related work . . . . 92
5.6 Summary . . . . 94
6 Conclusions 97 Appendices 101 A 3D environment modeling using heightmaps 103 A.1 Stereo images retrieved from a monocular camera . . . . 103
A.2 The Microsoft Kinect sensor . . . . 104
A.3 Quantitative and qualitative analysis . . . . 105
List of figures 107
List of tables 109
Bibliography 111
xvi