Bio-inspired emergent control of locomotion systems. /

1.1 The Central Pattern Generator (CPG) .. ... . . . 2
1.2 Locomotion control in hexapods ..... 4
1.3 Main topics of the work ............. . 6
2. CNN-based Central Pattern Generators 9
2.1 Introduction . ................................9
2.2 Brief overview on CNN architectures .. . . . . 12
2.3 The CPG neuron . . . . . . . . . . .. . 14
2.3.1 The synapse model . . 18
2.4 The CNN-based CPG . . . . .... .... . 21
2.4.1 RD-CNNs to design artificial locomotion patterns . 22
2.4.2 Guidelines for CNN-based CPG design . . . . . 24
2.5 Example: the caterpillar gait for hexapods . . . . . . 28
2.6 A spatio-temporal algorithm for controlling locomotion of a
hexapod robot . . . . . . . . . . . ....... . 32
2.7 Motor-neurons and inter-neurons . .. . . ... . . 37
3. CNN-based CPGs with sensory feedback and VLSI im-
plementation 43
3.1 Direction control ... . . ..... .. . 43
3.1.1 The CPG cell .. ...... 44
3.1.2 CPG with sensory feedback for direction control . 47
3.2 Feedback from ground contact sensors .. 50
3.2.1 Behavior of the CNN neuron driven by a periodic forc-
ing signal . . .............. 51
3.2.2 CPG with ground contact feedback: results .54
3.3 Reflex implementation ... . .. 57
3.3.1 The elevator reflex . . ... . . 57
3.3 2 The searching reflex . . .......... 58
3.4 Speed control .. . ...... 60
3.4.1 Speed of the gait .... ..... 61
3.4.2 Locomotion pattern . . ... .... . 63
3.5 The CNN-based CPG VLSI chip ....63
3.5.1 The hybrid approach ....... .. .. 63
3.5.2 The VLSI Circuit Design .64
3.5.3 Experimental results . .. . 66
4. Decentalized locomotion control 73
4.1 CNN-based decentralized control model ... 73
4.1.1 The decentralized control paradigm ... . 75
4.1.2 The CNN leg controller . . . ..... 77
4.1.3 The whole control system and results . . . 80
4 1.3.1 CNN Decentralized Control . .. 80
4.1.3.2 Choice of the parameters of the model 83
4 1.3 3 Robustness of the CNN Decentralized Con-
troller . . . . . . . . . . . . . . . . .. 86
4.2 Integrate-and-fire neurons and decentralized control . 88
S.2.1 The leg controller . ...... 90
4.2.1.1 Biological motivations . .... .. . 90
4.2.1.2 The integrate-and-fire neuron ...... 90
4.2.1.3 Scheme of the leg controller ... 91
4.2.1.4 The elevator reflex . . . . ... 95
4.2.2 The whole control scheme ...... 97
4.3 CPG and decentralized control . . ...... . 99
5. A gallery of bio-inspired robots 101
5.1 l,ampbot: A lamprey robot controlled by RD-CNN .. 101
5.2 MTA hexbot: a hexapod robot controlled by MTA-CNN .. 106
5 3 MTA hexbot II: a remote-controlled hexapod robot . .. 110
A5.4 MTA hexbot fII: a robot driven by the CNN-based CPG
VLSI chip ... . . . . .... . 112
6. High-level analog control: attitude control and Motor
Maps 115
6.1 CNN-based attitude control . . . .. ... . . 115
6.1.1 The CNN for gait control .. . .. . . 117
6.1 2 The attitude control CNN ... .. ... 119
6.1.3 Experimental tests ... . . 123
6 2 Motor Maps and attitude control . .. .. . .. 125
6.2.1 Motor Maps . .. .. 125
6.2.2 Motor Maps for Chaos Control . .. . . . . 129
6.2.3 Motor Maps for attitude control . . . . .. . 132
6.2.4 Motor Map-based attitude control in a simplified
biped model ......... . 136
6.3 Learning with Motor Maps . . . 142
7. High-level analog control: Turing patterns and autowaves 145
7.1 Reaction-Diffusion CNN ........ ..... 146
7.2 Navigation control based on Turing patterns . . . 148
7 2 1 Turing patterns and CNNs ................. 149
"7.2.2 From CNN patterns to action patterns ...... 151
7.2.3 Experimental Setup ......... 153
7.2.4 To probe further . . . . . ..... . . 154
7.3 Navigation control based on autowaves . . . . . 156
7.3.1 The CNN algorithm ... ..... . . . 157
7.3.2 Implementation on a roving robot and experimental
"results . . . . . . .. . .. . 160