Ted: an early order of business will be settle on graph engine, Neo4j and Tinkerpop seem like contenders but want to hear based on experiences
Joakim: Neo4j perhaps
proprietary in their graph query language, tinkerpop I think
supports multiple well
... what is the intent of this graph project?
Ulf: we decided at the f2f that
it would be useful to gather some vehicle data from fleets or
test vehicles and work it into VSS
... we will want to take into account GDPR and anonymization of
course
... we can then make it available on this server at MIT through
graph interface
... we can use this to demonstrate to others to help engage
them in this standard effort
Joakim: I have access to trip and
sensor data but key will be to strip any private
information
... there could be some interesting analytics performed on
it
Glenn: correct, we publicly
expose a substantial amount of data at Geotab for public
consumption that has gone through various scrubbing
... data.geotab.com
... we should define the privacy protocols. we worked with
Lothar who you may remember from the workshop on this
protocol
... we have had legal departments assess at universities we
collaborate with
Joakim: I guess we'll have to
write the translators to VSS
... Ulf's first responders use case was interesting
Ulf: all participating would need to be able to come up with these core signals to meet the use cases
Ted: Gen2 - web sockets and REST
(HTTP). ViWi has Graph and HTTP->Graph is logical
... sockets+http in-vehicle, http+graph in cloud. we can work
to combine with Ulf's Gen2 implementation and influence
spec
... something tangible for people to play with and see the
merits of standard
Joakim: if we decide on what signals, I can look to see if we can come up with a dataset to map
Ulf: I think we need to get
interested parties together and extend table on wiki to
identify what signals
... we need to formulate together, rules and policies for what
is an acceptable anonymization level
... I think Glenn mentioned we already figured out something at
Geotab we could share
... having a common if not standard set of anonymized signals
that can be exposed would be helpful
Joakim: did Prof Lothar Determan produce a document we could follow?
Glenn: yes, he took a
conservative approach based on GDPR (he is professor at US and
German universities) and California regulations
... separating time and space alone goes a long way. getting
permission from vehicle owners to participate provides consent
as well
... we have a document and policy under review that we want to
share
Joakim: besides just being online at W3C, we will want to get it out there as part of presentations
Glenn: I have one coming up at SAE/Government annual meeting in DC in January and I could provide an update on the project at that point
Ted: Speaking of governments and
data I have been invited to present at a EU Commission meeting on
Monday on telematics data governance but declined since it is too
short notice not to mention political
Singapore GovTech has invited me to contribute to draft on data
requirements for connected an autonomous vehicles as a MIT grant
project
I am participating next week in Canada National Research Council
AI4 Logistics (with
US DOT) on data requirements for regulators for national
transportation data interests
while I try to avoid politics and tell people we are working on a
technically flexible solution for the business and legislative
eventualities, I do feel there is a middle ground for the various
stakeholders
I want us to
define a nuanced solution (policy+best practices) that makes clear
what subset of data can be used by which third party app
in-vehicle, polling frequency, off-board to where and need for
independent security assessment before business decision to allow
it to run in a given OEM vehicle
I have outlined at a high level how this could work with various
OEM execs and it was received favorably as they understand it is
impossible for them to collect data in the cloud for every
possible need, sometimes in-vehicle data sampling and edge
computing will be necessary
I need to at least draft an outline for us and may be worth
discussing in this group along with our prototyping
Joakim: we could work on the best practices and policy work in parallel
Ted: it is an opportunity to define something before it is demanded (by regulators)
Joakim: I can show our legal the anonymization ideas from Geotab
Next steps: decide on graph engine
refine signals for use cases in Ulf's wiki
basic policy on anonymization and/or clear consent rules
identify who provide data to be scrubbed and translated to VSS
Ted: tech and best practices/policy work can in parallel. other next steps
Ulf provides:
[[High level view
Create a list over potential big data analytics use cases, including required signals.
Project decides on a few to focus on.
Project defines a common understanding on appropriate rules, policies, and practices related to what is required for data to be anonymous, not to breach GDPR, or other similar requirements.
Project participants selects an appropriate internal sets of data related to the selected use cases, and applies the practices from the previous step.
Project participants transforms the data set from their proprietary format to the VSS format.
Project participants upload the transformed data to the MIT server.
Project participants work together on using big data analytics tools on the aggregated data lake according to the selected use cases.
The project and its results are presented to the automotive community at conferences. ]]
Ted: we'll refine and get into wiki
Joakim: this is interesting and excited
Ulf: we all stand to learn from this
[adjourned]