Professional Documents
Culture Documents
a Lane Change
Mo#va#on:
to
control
and
minimize
the
higher
risks
associated
with
the
development
of
self
driving
cars,
which
are
specifically
designed
to
interact
with
humans
on
public
roads
Documentation
of
Validating
a
Lane
Change
• Safety
prac<ces
are
becoming
more
regulated
as
industries
adopt
a
standardized
set
of
prac<ces
for
designing
and
tes<ng
products.
ISO
26262
addresses
the
needs
for
an
automo<ve-‐specific
interna<onal
standard
that
focuses
on
safety
cri<cal
components.
• Valida<ng
a
lane
change
starts
very
early
in
the
design
process
with
assigning
and
documen<ng
an
Automo<ve
Safety
Integrity
Level
(ASIL),
the
key
component
of
the
ISO
26262,
to
the
lane
change
func<onality
of
the
vehicle
model
Documentation
of
Validating
a
Lane
Change
• Hardware
components
are
typically
qualified
by
tes<ng
the
part
in
a
variety
of
environmental
and
opera<onal
condi<ons.
The
test
results
are
then
analyzed
with
various
numerical
methods
and
presented
in
a
qualifica<on
report
along
with
the
tes<ng
procedure,
assump<ons,
and
input
criteria.
• Qualifying
so6ware
components
involves
ac<vi<es
such
as
defining
func<onal
requirements,
resource
usage,
and
predic<ng
soJware
behavior
in
failure
and
overload
situa<ons.
• Test
tool
qualifica#on
is
the
extensive
process
to
document
the
acceptable
results
of
the
test
tool
considering
the
limita<ons
built
into
the
test
tool.
Results
are
rela<ve
to
how
the
results
apply
to
real
world
driving
tests.
A
test
tool
used
for
self-‐driving
cars
is
a
simulated
world
were
human
designed
scenarios
can
be
run
through
all
or
parts
of
the
vehicle
model.
• “Proven
in
Use”
argument
is
simplified
documenta<on
to
ra<onalize
skipping
the
extensive
risk
analysis
at
the
beginning
of
the
valida<on
process.
Validating
a
Lane
Change
with
Simulation
Tools
• Unit
tests
examine
the
performance
of
individual
features
or
sub-‐features
of
the
whole
vehicle
model.
Results
from
these
tests
are
the
least
applicable
how
the
vehicle
model
might
perform
in
the
real
world.
• Situa#on-‐based
open-‐loop
tests
examine
the
performance
of
a
limited
number
of
interrelated
features
of
the
vehicle
model.
The
scenario
is
well
defined
and
actors
in
the
simula<on
react
to
the
vehicle
model
in
a
well
defined
manner.
These
tests
can
be
iterated
numerous
<mes
while
expec<ng
similarly
comparable
results.
• Scenario-‐based
closed-‐loop
tests
examine
the
performance
of
the
whole
vehicle
model
in
response
to
a
full
suite
of
simulated
sensor
inputs
that
mimic
the
interplay
of
individual
actors’
expected
behaviors
over
<me.
The
limita<ons
built
into
this
tool
are
the
expected
behaviors
programmed
into
individual
actors
which
makes
it
difficult
to
simulate
all
possible
edge
case
outcomes
from
a
single
scenario.
Validating
a
Lane
Change
with
Simulation
Tools
Validating
a
Lane
Change
with
Functional
Practice
• Results
from
closed
track
tests
are
easiest
to
quan<fy
where
most
variables
are
controlled.
These
tests
are
most
relevant
to
the
test
tool
and
simula<on
results,
but
do
not
best
indicate
how
the
vehicle
model
will
behave
in
all
scenarios.
• Controlled
variable
tests
on
public
roads
are
limited
to
the
subjec<ve
judgment
of
a
trained
expert
witnessing
the
vehicle
model
feature
being
tested.
While
very
few
variables
are
actually
controlled
in
these
tests,
the
real
world
contains
so
many
interrelated
situa<ons,
that
effec<vely
determining
acceptance
of
a
vehicle
model
feature
must
be
rela<ve
to
the
reac<on
of
human
actors
in
real
life
scenarios.
Key
to
determining
the
public’s
acceptance
of
a
vehicle
model
feature
is
the
observa<on
by
a
highly
empathe<c
expert
whose
goals
and
values
are
to
document
whether
an
interac<on
was
both
safe
and
effec<vely
courteous
considering
the
most
relevant
aspects
of
the
scenario.
Validating
a
Lane
Change
Resources
Referenced
• What
is
the
ISO
26262
Func<onal
Safety
Standard?
-‐
Na<onal
Instruments,
hVp://www.ni.com/white-‐paper/13647/en/
• Automated
Driving;
Safer
and
More
Efficient
Future
Driving,
2017,
Chapter
19
Tes<ng
and
Valida<ng
Tac<cal
Lane
Change
Behavior
Planning
for
Automated
Driving