Professional Documents
Culture Documents
Bartolo ZZ I
Bartolo ZZ I
Humanoid Robot
Chiara Bartolozzi
Our goal…
Smart Autonomous Systems
Event-Driven iCub
Vision Computation
Touch
Communication
Touch
ED Sensors — Touch
“Transient” taxel
ED Sensors — Touch
Sustained
Transient
• Capacitive Sensors
Tactile sensing
Taxel
digital converters
Off-the-shelf
RH
Taxel
Taxel Taxel Taxel
Taxel Taxel Taxel
Taxel Taxel
Taxel
Taxel ADC
ADC Taxel
Taxel ADC Taxel
Taxel
Artix7 FPGA
soft-core
Acquisition controller
Tactile hub
T
T H T
T
T H T
Events generator
Daisy
Asynchronous
H H chained
events
T T
Events arbiter
tactile
T T hubs
T H T
T T
T T T
T H T
T
T T
infrastructure
AS-AER body
Asynchronus
sensor
events
Robotic head
T
T
T H H T
T Events timestamping
ZynQ7000
T
platform
Events acquisition
T
T H T
T
T H T
Event based Sample based
RH Robotic head
ED Sensors — Touch
Noise Threshold - θ
Level Threshold - Δ
Fingertip - 12 taxels, #7 active
ED Sensors — iCub Vision
SENSORE
Figura 25 Dettagli geometrici e tolleranze ottiche per il sistema
Python CMOS
spazio-variante per robotica umanoide
ante attività ha interessato la progettazione di un obiettivo per l’occhio
iCub. Attualmente quest’ultimo impiega una lente grandangolare
coppiata ad un sensore matriciale del formato 1/3”, con una Back Focal
Spatial 1,3Mpxl
lto bassa. Dal momento che si vogliono disporre due sensori dietro la stessa
aumentare la BFL per accomodare un elemento di separazione (Beam-
fronte d’onda (Fig. 26).
Res 1/2”
Temporal 42 fps
Res (24ms)
AFOV 42°
ED Sensors — Vision
n t s
V !
E v e
e n Ct o n
o pio n ATIS t io CMOS
N o
V is m a m
g e
t !
e r fo r t re a n
I m a
p u U T l I n s S Spatial
i s io 240x304
o o m B o ra e n t V
Res 2/3”
N
t C m p E v i r e!d
d ap T e o u s s p
i m e Temporal 1μs
A l o i t i nu l I
d n T Res
x p o n t i c a
a n
E C log AFOV 61°
B i o
ED Vision — Optical Flow
ED Vision — Optical Flow
8 8
7 7
5 5
4 4
- spatial derivatives 3
2
3
1 1
1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9
Score
R negative 0
- corner: two large eigenvalues
R big -0.5
-1
-10 -5 0 5 10
Distance from Ground Truth (px)
ED Vision — Corner Detection
ED Vision — Corner Detection
3 3
2
Error < 2pxl 2
Error (px)
C1
120 C11 1 1
C2
100 C3
C12
C13
80 0 0
C8 10 25 40 55 70 85 100 0 15 30 45 60 75 90
y (px)
50%{
3x{
250 60
C7 4
75
0
0 20 40 60 80 100 120 200 90
x (px) 3
150
2
100
1
Vasco et al. IROS 2016 50
0 0
0 10 25 40 55 70 85 100 0 10 25 40 55 70 85 100
Speed (deg / s) Speed (deg / s)
ED Vision — Disparity and Vergence
LEFT RIGHT
Disparity:
Phase-Shift
Model
Vergence Control:
Proportional Velocity
ED Vision — Vergence Control
with normalization
dark environment
Thanks!