1. IOT: from a Prototype
to a Product
Lidwine Martinot
2. Outer shell:
Model it:
Blender, sketchup
Do it:
“Tôlerie plastique”(folded plastic)
Cheap small CNC
3D printer
HW:
Raspberry Pi, Beagle Bone, Arduino
Breadboards
Adafruit!
SW:
Linux
OpenCV, ROS, Qt
Stack Overflow!
Prototyping: Easy Peasy?
3. • There is no limit!
• Add a Fingerprint Scanner to Your
Garage Door Opener
• Keep Your Cats away from Furniture
with a Auto-Trigger Spray Gun (!)
• Mute Any Phrase You Want on Your
TV
https://lifehacker.com/top-10-kickass-arduino-projects-1747407543
How fancy can it be?
4. • Does the product solve an issue?: not the topic of today’s presentation but
the hope is yes…
• Look and feel, is my product attractive?
– Will customers be more likely to buy my spray gun if it is made out of precious
material and offered in various colors?
• What about the legal aspects
– Does my spray-gun have to be EMC compliant?
– What if my product accidentally kills some cats
• How reliable should my product be?
– How long a warranty should I offer?
• How to deploy it?
• How to maintain it?
• Etc etc etc
Let’s make a product then!
9. • Choose a designer who also has some ideas
about the functional aspects:
– How to insert the lock
– How to insert motors and LEDs in the robot
• Our designer also made choices for the prototype
but they had to be reviewed mainly for cost
reduction:
– Price of material
– Redesign to reduce processing costs
Challenges
10. • Choice of material:
– Aspect, properties (eg: in contact with water, sun ?), price, durability, not
independent from techno used to put it in shape
– Even with plastics: PA6, ABS, PC, ..
• Techno:
– Usinage, resin mould, injection moulding, 3D printing, thermoformage…
– Usinage: can we use a single-size drill?
– 3D printing: which aspect, “frittage”? Painted?
– Soft mould: up to 20 pieces
– Injection mould: infinite pieces
• but pricy to design and build the mould
• Needs a redesign of what has to be moulded
• Need to be sure not to have to change our model!
• Excel and Castorama have become a friends….
• Not just one set of techno/materials: one for small to medium
series and start from a blank sheet for bigger series
Challenges
Essai medium et
peinture
carrosserie
12. « Les dépouilles doivent être d’au
moins 0,5° pour éviter tout stress à
l’éjection de la pièce. Le plastique
injecté suit le chemin le moins
contraignant lorsqu’il entre dans le
moule, donc la matière doit pénétrer
dans la partie où la section est la plus
épaisse et finir dans les plus fines. Pour
un meilleur résultat, l’épaisseur de
paroi est uniforme ou varie de moins
de 10 %. »
Productization: Know your limits
http://eduscol.education.fr/sti/sites/eduscol.education.fr.sti/files/ressources/techniques/4721/4721-185-p30.pdf
Moulding a pedal:
We need a specialist!
15. Challenges• For a product requiring a radio:
– For the prototype: doesn’t matter as long as it doesn’t prevent
from making a demo. WiFi + tethering on a phone is good enough
for example
– For the product: Which technology?
• Depends on the coverage vs customers targeted
• What is the cost of the radio chipset?
• Can we easily find boards which integrate it already? Are they certified?
• What about subscription fees?
https://raed.it/blog/iot-network-sigfox-vs-lora/
16. Reliability
• Reliability: be “agile”
– Our test bench showed a huge variation in lifetime
for the 1st set of motors we had selected:
• How long do they have to last?
• How much does it cost to have someone intervene for
changing parts every year?
• Eventually: we selected new winners
– Never seen before failure during the Vivatech event
in 2017, after they had issues with the AC:
• Had to change a component on the card
• It’s hard for a small company to simulate all possible
environments the products will be in: take some risks
• Certifications: country-dependent
17. Cost
• Cost reduction effort
– Initially: flexibility on the board
– Once fully tested: fix the HW design
– More optimization: choice of components
• For productization we again decided to go in 2
phases:
– small/medium series before a bigger cost-reduction
effort for bigger volumes
19. The robot can
play a demo
Failure rate:10-1
Next phase:
SW redesigned to be more stable
but also to ease the dev of new
demos
Failure rate: 10-2
Nirvana!
20. Challenges
• Reliability… again
– Initially: failed to start once every 10 On/Off: we don’t even pay
attention to such things on a proto!
– Then… freeze once every 2 weeks: ok or not? No
• Addicted to Stack Overflow, and one day, you become a
contributor!
– There is always someone who has had the same issue. Amazing, but
as the issue becomes tricky…
– There is not always an answer provided!
• Increased Expertise for a more reliable product:
– Need to manage your own Linux image ? (build root dark side)
21. SCM and Maintenance
• SW Configuration Maintenance:
which version is where? Which one
needs which upgrade?
– Too obvious to mention it? Even for a
prototype SCM is a must
– But: can’t afford a QA team from the
very beginning
• OTA Update:
– Another side of SW: the maintenance
server
22. • Maintenance on a
prototype:
– drive to the customer’s,
fix it
• Needless to say…
doesn’t apply to the
product
25. • We haven’t been reading the user manual to
build that box, don’t expect our customers to
read it to use our product…
• Lesson learned: Make it simple for them to
use
• And now we have to productize the
packaging
Challenges
28. • KiOne is a tough product to productize but some
of our challenges apply to any IOT device
• Have fun, don’t hesitate to make a prototype
• Don’t wait for the perfect product to start
exposing it
• Yet…. It’s easy to underestimate the efforts to go
from a prototype to a product, even when we
think we know!
To conclude