T E L E M E T R Y
enhancing & unlocking new potential in OSM
Slide 2
Slide 2 text
OV E RV I E W
what is telemetry?
how does Mapbox handle telemetry data?
what we can calculate from telemetry data?
what are the use cases?
traffic
Slide 3
Slide 3 text
T E L E M E T RY
any data collected from a remote device
[lng, lat, alt, timestamp]
telemetry
pipeline
clock
GPS sensor
+
Slide 4
Slide 4 text
“probe”
[x, y, z, t]
P RO B E VS . T R AC E
Slide 5
Slide 5 text
“trace”
[ [x, y, z, t], [x, y, z, t], [x, y, z, t], [x, y, z, t] ]
P RO B E VS . T R AC E
Slide 6
Slide 6 text
A N O N Y M I S AT I O N
Slide 7
Slide 7 text
A N O N Y M I S AT I O N
Slide 8
Slide 8 text
A N O N Y M I S AT I O N
Slide 9
Slide 9 text
A N O N Y M I S AT I O N
Slide 10
Slide 10 text
P R I VAC Y
only store anonymised data
data is encrypted using industry best practice
clear boundaries around who has access & for what reason
audit-able access logs
more info – https://www.mapbox.com/telemetry/
Slide 11
Slide 11 text
“probe”
[x, y, z, t]
“trace”
[ [x, y, z, t], [x, y, z, t], [x, y, z, t], [x, y, z, t] ]
Slide 12
Slide 12 text
B E A R I N G
Slide 13
Slide 13 text
speed =
distance
duration
S P E E D
d
∆t
Slide 14
Slide 14 text
acceleration =
∆speed
duration
AC C E L E R AT I O N
Slide 15
Slide 15 text
U S E C A S E S
Slide 16
Slide 16 text
VO LU M E
from probe density
missing streets
satellite imagery alignment
health of satellite imagery
Slide 17
Slide 17 text
VO LU M E
from probe density
missing streets
satellite imagery alignment
health of satellite imagery
Slide 18
Slide 18 text
VO LU M E
from probe density
missing streets
satellite imagery alignment
health of satellite imagery
telemetry
density
raster
satellite
imagery
edge detection
edge detection
vertical
offset
horizontal offset
*
Slide 19
Slide 19 text
VO LU M E
from probe density
missing streets
satellite imagery alignment
health of satellite imagery
telemetry
density
poor imagery quality
Slide 20
Slide 20 text
M A N E U V E R S
from trace density
restrictions (turns, one-ways)
high-precision trajectory of
lane centreline for ADAS
traces present
absence of traces
Slide 21
Slide 21 text
M A N E U V E R S
from trace density
restrictions (turns, one-ways)
high-precision trajectory of
lane centreline for ADAS
Slide 22
Slide 22 text
T R A F F I C
from speed, bearing, acceleration
road congestion
traffic-aware navigation
Slide 23
Slide 23 text
T R A F F I C
from speed, bearing, acceleration
road congestion
traffic-aware navigation
Slide 24
Slide 24 text
T R A F F I C
Slide 25
Slide 25 text
W H Y T R A F F I C
origin destination
estimated
trip duration
(ETA)
date-time
traffic-powered OSRM
OSRM public demo server
Slide 26
Slide 26 text
LUA P RO F I L E
rule-based
static
100% road coverage
Slide 27
Slide 27 text
T R A F F I C P RO F I L E
telemetry-based
dynamic
coverage fluctuates
(never 100%)
Slide 28
Slide 28 text
N O I S E R E M OVA L
duration
distance
speed
bearing
deceleration
acceleration
modality
Slide 29
Slide 29 text
P RO B E M ATC H
distance from road
bearing from road
highway hierarchy
Slide 30
Slide 30 text
P RO B E M ATC H
distance from road
bearing from road
highway hierarchy
Slide 31
Slide 31 text
P RO B E M ATC H
distance from road
bearing from road
highway hierarchy
Slide 32
Slide 32 text
P RO B E M ATC H
distance from road
bearing from road
highway hierarchy
higher rank
lower rank
Slide 33
Slide 33 text
P RO B E M ATC H
tertiary road
distance from road
bearing from road
highway hierarchy
Slide 34
Slide 34 text
N O R M A L I S E D
G R A P H
traffic data uses
split at
road class change
intersection
merge otherwise.
Slide 35
Slide 35 text
N O R M A L I S E D
G R A P H
traffic data uses
split at
road class change
intersection
merge otherwise.
Slide 36
Slide 36 text
N O R M A L I S E D
G R A P H
traffic data uses
split at
road class change
intersection
merge otherwise.
Slide 37
Slide 37 text
0 20 40 60 80 100 120 140 160
Speed [kph]
Count [-]
W H I C H S P E E D
Slide 38
Slide 38 text
0 20 40 60 80 100 120 140 160
Count [-]
W H I C H S P E E D
primary
secondary
tertiary
motorway
trunk
links
alleys
service roads
…etc.
Speed [kph]
Slide 39
Slide 39 text
N O R M A L I S E D
G R A P H
traffic data uses
Slide 40
Slide 40 text
E D G E - BA S E D
G R A P H
OSRM uses
transform
maneuver → edge
street → node
Slide 41
Slide 41 text
segments
N O R M A L I S E D
G R A P H
traffic data uses
E D G E - BA S E D
G R A P H
OSRM uses
Slide 42
Slide 42 text
segments
N O R M A L I S E D
G R A P H
traffic data uses
E D G E - BA S E D
G R A P H
OSRM uses
T R A F F I C
M A P S
I N C I D E N T S
T U R N
P E N A LT I E S
R E A LT I M E
T R A F F I C
VO LU M E
Slide 43
Slide 43 text
T E L E M E T RY
ETA, error percentile, etc.
layer against
road network &
attributes
road network &
attributes
better, fresh satellite imagery
where to
focus
(proactive)
what is
suspicious
(proactive)
where/what
is an issue
(reactive)
where imagery
is outdated
DATA
T E A M
DATA
T E A M
S AT E L L I T E
T E A M
traffic