This keynote presentation showcases through concrete examples why modeling long-term usage of automotive infotainment should begin as early as possible in the development phase, rather than leaving issue detection as a concern for test fleets and end-users.

1. The new era of
software quality
awareness.
FROM GUESSTIMATES TO FACTIMATES

2. Discover hard-to-find defects before your end-users do.
Our technologies and services provide a prioritized summary of software stability,
performance, power consumption, and functionality levels at one sight.
Automotive Infotainment Smart and Medical Devices
Profilence Analytics

3. So you passed xTS tests!
But is your system stable?
THE LONG, HARD ROAD TO MARKET READINESS

4. Functional and compatibility tests don’t catch defects that occur in long-term use.
Your Android Automotive system might pass xTS tests with flying colors and yet fail in daily use.
There was a report from the field saying “infotainment crashes,” with no
information about how it happened and under what conditions. That’s when
the devs try to repeat the problem and are working during weekends. That
time is away from the family, and other projects will also stop there.
Nicolas Gattario, Haleytek
“
Issues may go unnoticed until discovered by end-users

5. Memory leaks are a typical finding in stability analysis. A SW
component’s memory consumption goes up until the component
or the entire system crashes.
This is an example of an AAOS-based infotainment system that
crashed constantly after 5 hours of use because of a memory leak
in one of the platform SW components.
Case 1: memory leaks
We had no idea such a defect existed. We had
never seen it in our internal testing!
Undisclosed infotainment vendor
“

6. Instead of shutting down, most infotainment systems suspend to
RAM when a vehicle is turned off, and resume their operations
when the vehicle is turned on again.
As a result, the system’s total uptime continues growing and long-
term stability issues start to occur.
Also, there are often issues with the system not resuming its state
properly, leading to e.g. Bluetooth instabilities:
Case 2: Suspend-to-RAM (sleep)
We had customers complaining
that their smartphones couldn’t
connect to the infotainment
system after the vehicle was
turned on.
Undisclosed OEM
“
pid: 24659, tid: 24763, name: bt_hci_thread >>> com.android.bluetooth <<<
uid: 1101002
signal 6 (SIGABRT), code -1 (SI_QUEUE), fault addr --------
backtrace:
#00 pc 000000000005a4e8 /apex/com.android.runtime/lib64/bionic/libc.so
#01 pc 000000000005d472 /apex/com.android.runtime/lib64/bionic/libc.so
#02 pc 00000000002bd451 /system/lib64/libbluetooth.so (hci_timeout_abort()+
. . .
15:11:38.132 Time to park the car
15:11:38.141 suspendForSeconds(180)
15:11:38.960 Sending STR enable command
15:13:19.188 Car suspended to RAM
15:13:19.189 Sleeping now for 159s
15:15:58.191 Resuming from suspend
15:15:58.202 Sending STR disable command
After
resuming

7. Stability problems don’t always manifest as crashes.
Sometimes, they might lead to an ever-slowing down system
that is still functionally ok, but unpleasant to use.
This is an example of an application that is slowing down
more and more each time it is started.
Case 3: Performance degradation
The UI felt somewhat sluggish,
but we didn’t know why!
Undisclosed OEM
“

8. Some defects occur only under specific, sometimes rare
circumstances. Root cause might not be in infotainment.
Such issues are extremely hard to reproduce. In these cases,
it’s vital to have data from the hypervisor, ECUs and other
critical entities, not just from AAOS.
In this hard reset example, the logs are from the AAOS host
module:
Case 4: Random, occasional crashes
We encountered some problems that
were very difficult to reproduce with our
standard testing procedures.
Undisclosed OEM
“

9. OUR QA SERVICES
We help you
reach market readiness

10. Known
issues
With Profilence
A typical case
Development phase After-market phase
Fleet trials
We accelerate:
§ Defect discovery times
§ Root cause analysis
§ Verification cycles of fixes
Reach high software quality
before fleet trials by
modeling long-term use in
R&D laboratories.
Avoid quality regression in
OTA updates.
Find, Fix, and Release faster

11. Our QA services support you from day 1
All services are provided by our QA experts and empowered by in-house QA technologies

12. Data from lab modelling and fleet trials are
consolidated into a single QA dashboard,
that shows:
• Stability trends between SW versions
• Performance trends between SW
versions
• Energy consumption trends between
SW versions
• New and known issues categorized
• Unique issues per SW version
• Detailed analysis of issues found
AS A RESULT:
• Management gets a 360° view of
current quality levels, issues and trends
• All QA data is available for OEM
ecosystem partners transparently
• Shorter QA cycles
• Rapid issue discovery, fixing and
verification
• Development processes become faster
and more agile
Real-time dashboard Keep project management, dev teams and QA in sync 24/7

13. The impact
You have full understanding of
the current quality level,
trends, and the most pressing
issues.
Quality awareness
Defects’ root causes are
known; engineering teams
focus on fixing them instead
of treating symptoms.
Root causes
Vast amounts of debug data
is available; teams don’t have
to reproduce issues to
understand them.
Debug data
Stability analysis is done in
laboratories and results are
available in days. Teams don’t
need to wait for data from fleet
tests.
Reliance on fleet tests
The analysis models long-
term usage and stability
issues are discovered early.
No more negative feedback
from your end-users.
Long-term usage
Unbiased analysis results and
debug data are available to all
teams and value chain
partners. Everyone is on the
same page.
Transparency
The risk of defects going
unnoticed until discovered by
end-users has been greatly
reduced. Projects stay on
schedule.
Risk control
QA costs are reduced and
can be estimated accurately;
budgets won’t get out of
control.
Cost-efficiency

14. Accelerate your data-driven
software development with our
world-changing methodology.
Profilence’s methodologies find the root causes of software
defects. The most advanced platform gives real-time
feedback to the whole OEM ecosystem.
FROM GUESSTIMATES TO FACTIMATES