Josh Shinavier - Uber, Ted Guild - W3C
Ted: First and foremost let me
welcome you, thank you for coming. I want to also thank Uber for
hosting and Geotab as our sponsor
... in email as you probably saw there was a stir over the terms
of the visitor's agreement
This is a common practice for tech and
other companies. NDA are not conducive to open standards
development and I am generally discouraged by W3C and MIT to
sign
... Uber understands these are open proceedings, minutes will be public and they and
everyone else are discouraged from confidential information in
this workshop
... If there is anyone with strong objection or hesitation from
proceeding under these conditions please speak up and we can
discuss possibility of changing venues, otherwise let's proceed
and appreciate our hosts' hospitality
[No objections raised]
[Josh introduces himself adds additional clarity on visitor's agreement]
[logistics for realtime minuting on IRC, making off the record comments, #w3ctransportation channel and tag, remote participation, facilities]
W3C Workshop on Data Models for Transportation[slides]
Josh: I wanted to tell you a little
bit about about this effort about data standardization, because I
think it's kind of analogous to what we're doing here [at this
workshop]
... So we started out about a year and a half ago with the realization that schemas are very fragmented within Uber.
Josh: current problems
slide (myriad of datasets, schemas...)
... Future problems slide (
... We are undertaking schema standardization here within Uber
(geospatial, time, sensor, payment...)
Josh: we have a good
notion of identity for a user but not a central model for
one
… we will have same
problem ingesting external data resources
<ted> Metadata vocabularies: datasets, services, lineage, privacy, SLAs, deletion and retention
Josh: Data model standardization @Uber
<ted> Entities and relationships
Josh: we released a
paper called algebraic propery graphs
… schema review is an
issue internally for us, need to incentivize schema developers
and coordinate beyond the company
Josh: Driving adoption slide (schema evolution, review and style challenges)
Jason Thomas (HERE): The model mapping, is there integration involved or persistence involved?
Josh: That typically involves
translators so no latency
... For the schemas the typical example is developing in
language A and definitions are in B and how do you do
that
... We use translators for that.
Amir Sayegh - Geotab [slides]
Amir: I'm with the Analytics part
of Geotab.
... There are 3 things to talk about.
... Who is Geotab
... Why we're working on AI
... Why it's important
... What is needed to keep data open
... There is a call to action to help do that.
... I joined Geotab a year ago from academia, now working on
Analytics and AI
... When looking at GeoTab, we're an overnight success 20 years
in the making
... we specialize in Fleet Tracking, a family owned
business
... we're a sole provider for the US Govt and the state of
CA
... We have a model where we go through resellers and
marketers
... We're out of Ontario, 30m outside of Toronto.
... A World Leader slide
... Geotab Technology slide
... The magic is the GO device we invented, we connect to the
ODB Port and we have a SAS model that lets you manage your
fleet
... We're connecting all the engine signals
... The device has an accelerometer that let's us pick things
up.
... The device has a GPS and cellular connection so it collects
all the info needed.
... We can verify compliance with government heavy vehicle
regulations and company policies, do they back in to spaces, fuel
economy, etc.
... Open Platform slide
... The overarching approach is to be open and
transparent
... in addition to providing all the data to a customer we can
also provide 3rd party data
... the Geotab device serves as a gateway.
... We've had a marketplace approach allowing 3rd parties to
build services off of our data.
... Our challenge to date is that every vehicle has a different
communication on the CAN bus.
...Amir: Turning Raw Data into Sensed Data slide
... There are ways to correlate data from raw signals to turn
them into sense data.
Amir: Vehicles Become Sensors on
Wheels
... In all vehicles you see temperatures so you can turn every
vehicle into a weather station.
... When we see a vehicle with a lot of hard stops it may be a
fault of the road.
Amir: Big Data at Geotab
slide
... The power of our data is you could map almost every road in
US and Canada within a day.
... Innovationg with Big Data slide
... We launched our public portal with data available
... Telematics Data can Transorm the Commuinty slide
[he reviews applications that the data can facilitate]
Amir: . . . .while Respecting
Privacy and Ethics slide
... We have a team that looks to make sure we're respecting
privacy.
... We have to worry about unexpected consequences of our
data.
... We've built in a "motivated intruder" model to look at data
before we anonymize them.
... Fleet Insights Through MVP Progam slide
... There is a lot of data and we don't know what we don't
know
... We look at our own customer information like Fuel
economy
... We can use that to guide the customer on vehicle type and
driver habits
... Smart Cities Slide
... Safety Using Big Data slide section
... Data Enables Safety Understanding
... We typically talk about data, and when talking about it we
need to understand the why something is happening.
... In order to do that we need data from the vehicle, were the
wipers on, were your flashers on, etc.
... We call this external data.
... There is contextural data. Everything that's happening with
the driver.
... We had a client that gave us 1700 accidents. One of the
things we saw was the number of accidents that hit
animals.
... They had to put things on their long-haul vehicles to
address it.
... Vehicle and Telematics Data slide
... You need to benchmark you against you as well as you
against all drivers.
... We need all data sets from the vehicle to provide the best
answers.
... We have cases where experienced drivers are running out of
gas which is a hazard.
... We need to understand why this happens.
... Example: Fleet Safety Policy Design slide
... Sensor Data + Big Data _AI = Safer City slide
... Supporting NYC Vision Zero slide
... Public Safety - Hurricane Dorian Impact
... Don't underestimate the power of a global sensing
network.
... Current Data Challenges slide
... We've heard things that concern us about some data
restrictions
... I'd like us to think about the data silos we have and the
value of sharing.
... "Only "hack" today is reverse engineering; concersn loom
over reduced access"
... Current Work by Geotab: Data Accessibiity Index Project
slide
... The 5 Step Methodology slide
... We started small, applied exclusions, etc.
... Current Output (Sept 2019)
... We looked at 1M vehicles, in 94% we could get to all the
data
... Across our base it was only 60%
... Magic Quadrant: Sample slide
... A Final Note: Data "Value Chain" slide
... We need to pool our data to be more successful.
... The value comes from what you do with the data.
... Call for Action! slide
... There are hundreds of applications that can be driven by
this data.
... Knowing we're sitting on these and unable to extract the
value we need to drive a culture of openness.
David Mordecai (NYU): What challenges are you facing
around OEMs and their warranties as well as challenges from their
firewalls being added to OBDII?
... It's becoming very limited on what
comes through there.
Amir: Specifically there is
US legislation (right to repair) that varies around the ability to void
warranties
...There's rumblings of that in other
jurisdictions. But again, I'm not the expert.
...this is why
we are creating an accessibility index
Amir: The ability to access data is certainly going down.
Mark Fox (UToronto): what is available at data.geotab.com?
Amir: will talk about it more in a later presentation
Mark: does this contribute to pic alert(?) from national center for atmospheric research?
Amir: no but we are actively looking for establishing such arrangements
DavidM: Do you have a view around
data fusion and data senses that provide you reliable
data?
... Seems you need to have some control over that.
Amir: I'll give you two
obsticales. One is there is no standard for safe sharing of
data.
... The second is the bias you're talking about.
... Geotab has a bias for commercial vehicles.
... We need to have a way to "de bias" these datasets.
... We get asked about extrapolating data and that depends on
the specific analysis.
Jenny Huang (IFOSS): I have two
questions. At the end of your talk you mentioned using private data and
the value to drive AI and do good.
... There is an initiative from the data collaboratives around
this and wonder if you are aware of them.
... My second question is from W3C in general, I think we're
doing the same analysis in the telco sector, the motivation is the
same to adequately anonymize data, handle privacy implications,
etc.
... I think the standardization is key. Where is the place to
do that. Telcom needs this too.
... I think we can work together vs. having things in
silos.
Amir: I have gone done the same path. Telcom data isn't as accurate but having one standard is key.
Alan Bird (W3C): couple things,
we're in the process of forming a new group, the scope is for
Telco needs and that can
be taken into consideration
… nearby is the results
from a Graph workshop we held in May in Munich and work is
starting there too. not industry specific but seeking verticals
to coordinate on their data exchange needs
Jenny: I think some of the data is real data that people would pay for.
Jay Hum - Autonomic [slides]
Jay: Everything around us probably came
here in a shipping container.
... simple shipping containers lacking technology disrupted the
transportation industry and economies internationally.
… tracking a container
from Malaysia to a LA dock, that is then brought by a truck to rail and shows
up 28 days later at BestBuy
… a ship this size can
unload 1k containers in 24h and move on to the next port
… the orchestration has
so many logistical elements it is surprising how in sync they
are
… how did this happen?
the most important development was containerization,
standardized sized containers, the cranes, rail and truck
trailers to accommodate them
… the value of the data
depends on the customers' need, can be accelerated through
common standards
… what we are building
at AU is TMC - Transportation Mobility Cloud
… it is focused on
transportation and mobility, an open cloud based API
platform
… Amir was showing
value of raw data compared to analysis results, we are
somewhere in the middle
… we are part of the
Ford motor company, taking data of eg F150s
… we are creating an
open cloud based platform
Jay: we're already looking to
handle other modes of transportation, bikes, scooters, drones
etc. the metrics and telemetry apply to many modes
... we are normalizing, ingesting and processing in order to
expose a suite of APIs
... allow for people to create data rich applications
... we can also go the other way, through an external
application communicating through our cloud API down to the
vehicle to eg unlock a car
... power and value of what we are doing with TMC
... we are embedding connected vehicle data from 4 different
devices and even then we have different protocols, this is
within one auto manufacturer
... we are trying to push our own protocol
Carlos Garcia (Autonomics): the various representations of the data is crazy as well
Jay: learning to deal with a myriad
of acros within auto was "fun"
... we have own view of a data dictionary, it is just a
collection of words
... we have three main elements: signal, indicator and
event
indicator - on or off
Jay: amoung those 4 devices and
protocols there are 400+ elements
... our job is to try to create an open cloud based
platform
... it is a challenge to represent their very specific data,
how to be able to expose it for more wide use
... we are looking at treating these elements as first class
citizens so we can decouple the data from the device
(vehicle)
... eliminate device specific or bespoke meaning
... move from device centric to data centric. data should be
able to stand on its own and you can then get more
insight
... insight has to be actionable
... we want people to leverage this data and create
solutions
... every oem has their own data dictionary, variations even
for their different networks (eg raw signals on CAN bus)
... we want to make the data more actionable and insightful.
for door lock example we want to specify which door, fobs at
present do not provide that
... hard acceleration alert event. it comes with two CAN
signals. we have taken it a step further creating a triggering
condition to get more context about why this event
occured
... context becomes very important. hard acceleration might be
a result of a bump from another vehicle instead of someone
stomping on gas pedal
... heading can be a good indicator along with what the pedal
actuator is telling us
... moving actions from device to cloud, allows you to extend
actions to the data
... we are trying to extract more value from the data beyond
the raw bits
... faster development and time to market
... suite of developers can get up and running quicker and
easier leveraging these apis
... high level view on a couple apis. streaming data
representing last known state, close to realtime. historical or
time series to do trend analysis and predictive
maintenance
... vehicle meta data - year, make, model...
... rental cars will expect all of you to fill up before you
return, often people try to see what they can get away with or
fill ahead of their return
... the visual check instead of checking data is
insufficient
Adnan: some OEM are making that data available in detail already
Jay: incidentally Geotab is using
our API, helping with the data
... we have more APIs geared towards Group and Vehicle basis
for quick aggregation and more complex queries
... one of our premium adds is our Geofence API and
service
... many use cases where those are useful for fleets and end
consumers albeit a bit big brotherish
David Mordecai: in your topped off tank for rental vehicles use case, various factors within the vehicle and at fueling provider can lead to values being off
Jay: yeah it is not an exact science, shapes of the tanks vary, can be parked at an angle etc
DavidM: if you do not share the information as a basis for pricing decision with accuracy it can be problematic
Dick Roy: hard braking events like you described does not provide the information you listed, does it go out to the cloud and then instruction come back to collect the other data points?
Carlos: some already send the
information contextualized, the event triggers the collection of
other data points
... we also will make the inference where we can
Dick: you said you are providing open access to this data from the cloud, can I get it for free?
Carlos: no
Dick: but what if it is my car?
[discussion of data rights for vehicle owner, deferred to tomorrow morning]
DavidM: hard acceleration and braking could be good driving avoiding an accident, there can be algorithmic bias
[agreement on data based assertions flawed from missing information]
Luis Daniel Ibáñez - University of Southampton [slides]
Luis: QROWD project is a
public/private project involving three universities, two large
companies
... various municipalities as well. EU commission funds things
deemed of interest
... in case of QROWD is how to include humans in the value
chain
Luis: first problem is survival,
dealing with CO2
... according to EU agency, cars are the main culprit with more
than 40% of carbon emissions and within that mostly light
vehicles
... this is a good argument for multi-modal adoption
... next is traffic congestion reduction. average person looses
2 to 4 days per year in traffic, this graphic shows it across
different cities
... we work extensively with EU commission, they ask us
questions and get input. Policies are Coming
... Europes answer to emeision reduction challenge is the shift
to low emission mobility
... transformation land plans to be evolved for different
regions
... Paris banned cars in city centre one Sunday a month. they
were not prepared to deal with it and handled complaints in
part with free Metro access for the day
... NY, Oslo, Chengdu, Copenhagen all have similar but varying
models
... two obvious solutions, low emission vehicles and
multi-modal travel. we are seeing a mix of the two
... in US seeing more low emission vehicle focus
... to balance emissions, traffic, travel time and costs
... for us we have been studying needs of going from A to
B
... I may take an Uber to CalTrain and from the end station a
scooter
... we need some sort of data standardization to accomplish
this
... there are various one off efforts by companies and
researchers
... for public sector, accurate modal share data. a
municipality is particularly interested in traffic congestion
information and forming data driven decisions
... Geotab for example is fleet focused. Need to get from eg
TomTom consumer vehicle information to combine the
datasets
... also interested in journey planning. consent a bit
issue
... for private sector, customer engagement, help with their
mobility
... what we need is to enable data sharing spaces, in context
of urban environment and also ad-hoc sharings p2p
agreements
... how are we going to do that data sharing?
... include the travelers for access to their data, feedback
and their preferences
... challenges we see are vehicle heterogeneity from
traditional bike to connected vehicle
... combining efforts and coordinating with C-ITS, ISO TC204,
plus what private companies are doing
... trust, accountability, data accuracy and validation
... what should be our role?
... should we try to define these topc classes? Vehicle, Road,
Trip, Traveler or we going to go from the various current sdo
proposed representations
... we need a "Web of Vehicles"
DavidM: what is your point on the standards vs definitions?
Luis: we want to avoid confusion of competing efforts. should we stay high level, identifying what is needed or start from the detailed solutions being started
Adnan: why won't it work?
Ghislain: @@x?
Luis: we are working on a citizen data model, need to reflect decisions made by individuals
Dick_Roy: who is we, W3C?
Luis: QROWD
Dick: where do you see the standards piece and what should it do?
Ken_Vaughn: I will be presenting
on ISO TC 204. I view your top line as a logical data
model
... if someone wants to develop something in more detail that
would be below the line
Jay: I want to be able to link up the different modes of travel and deal with this pain every day
Brent Massey - Ridar Systems [slides]
Brent: we are lacking the
connectivity between modes of transportation and pedestrians
and relying on line of sight visibility
... our goal was to provide a safety element by providing data
to drivers about location of bicycles
... motivation for vehicle operator includes reduction in
insurance
... goal is to get down towards zero accidents
... we need to get the end uses, the daily commutters on
board
... we can cover this as well during our breakout
... incentivize first but move eventually to enfocement
... adaptation of infrastructure and streets is a challenge,
cannot remove needed parking spaces in all cases to create bike
lanes
... what we are proposing is a passive safety system that can
be included in ADAS systems beyond line of sight
... we want to see this get including into head unit
... rider sets it on their phone using our app, uses gps,
accelerometer etc and be able to communicate trajectory
... turns the bike into a smart beacon
... two wheelers want to be seen
... could not help but notice the amount of visibility attire
and lights people are using around here
... it is a cloud based cellular but can also be intergrated.
cyclists do not have enough power on their phones, maybe tape
into batter of ebike and other e-mobility modes
... our data is anonimized. drivers get a notification based on
location, direction and speed and when it predicts it is
approaching a common point, alert the vehicle driver about 5s
before they intersect
... this allows you to see in front of the vehicle ahead of
you
... we use cellular, gps. loss of signal will be communicated
to rider
... another use case is a high volume of cyclists in the area
and switch to a general instead of individual notification
until bikes thin out
... the data warehouse is availabe and interested in sharing
strategically
Dick_Roy: you have multiple
transponders sending to the cloud information at high
rates
... this is going to take cellular system to its knees
Brent: we use far less data than
eg Google maps
... work is down on the cloud
Dick: worried more about network
capacity
... doesn't scale when it comes to individual pedestrians in a
major city
DavidM: this is a commendable function but creating a false record of data
Brent: we understand the limitations and level of accuracy
DavidM: understood and appreciated
Martin Kurze (DT): 5G will help
solve this. a part of 5G standards includes precise positioning
for this sort of scenario
... you said you don't care about privacy because you anonymize
but that is hard to do with precise data...
Brent: a balance, definitely
Mark Fox (UT): I do not see a need to go up to the cloud and centralize processing. this can be done with more proximity based emitter on bicycle and vehicles able to interpret
Brent: they require more power
and components than people have available on two wheelers at present
... also heat dissipation. you don't want to draw down
charge
Dick: CA did this already as a proof of concept. DSRC can receive notices from pedestrian's phone
Josh: from Zoom Michelle points out there is already effort as part of C-ITS
Robert Rittmuller - US DOT Volpe [slides]
RJ: I work at the US DOT Volpe
research center in Cambridge, Massachusetts
... I am here to talk about data models for safety, research,
cybersecurity and future vision and topics we want people to
think about
... fundamentally 99% of our work is tied to safety, constantly
using that lens
... someone earlier mentioned vehicle crash research. we were
involved in minimum uniform crash criteria
... what this gives us is same terminoligies on that
dataset
... one project I work on a connected vehicle project and we
are working on schemas and data models to bring BSM into
common
... Basic Safety Message
... in context of AV think of the data needed for them
... Work Zone Data Exchange WZDx is a real problem area for AV
to navigate
... we have a github repository people can interact with and
encouraging people (state gov) to adopt it for those who can
consume it
... we are doing pure research and trying to extract insights
for researchers. I am invovled in the AI group at Volpe
... SHRP2 NDS - goal of the project is to a video data set to
implement to do naturalistic driving study
... we were asked to annotate based on computer vision
techniques
... Phase I is complete and starting II
... one of the things important to notice is we are planning to
make these available online, open source
... those using the SHRP2 dataset can search for these
annotations
... a little more interesting is a project for FAA - cockpit
audio transcription
... when we try to develop the language and audio model, none
of the voice recognition can read it. again looking to open
source but FAA is hesitating
... we want to break down silos and make data available. a
transcription of cockpit conversations internal and with air
traffic control is valuable and useful
... we delve into legacy transportation systems, outdated
infrastructure
... we wanted to do AI assessment of these legacy systems
... Government Telematics Security Conformance Acceptance
(GTSCA)
... we are looking for what bad actors to do to individual
vehicles and fleets
... in-vehicle ethernet is starting to emerge
... we are very concerned about data privacy. we want to avoid
regulation if we can help it but it is becoming an issue
already
... a MIT research group geolocated every single crash
accurately after we tried to anonymize it
... you may think you have data privacy but you don't
... if a vehicle is equiped with a telematics device. at defcon
someone did simple sqli to get into a remote start device
... you need to consider what data elements can be
misused
... we are also concerned with IVI, ELD, in-vehicle remote
systems diagnostics
... the infotainment systems are quite vulnerable
... I found an infotainment system that when paired with my
phone started leveraging it for data
... we are focusing more on ADAS than AV at this point
... what we are finding out these systems vary widely
DavidM: they need to be tested in the real world
RJ: correct, we are asking for that to affirm safety
RJ: fundamental problem is bespoke solutions, we need to look at real world trips and be able to normalize
Mark: why isn't the government coming up with a common simulator?
RJ: [company] invested over $1B
in theirs, we can't do that
... how do we know if it is accurate enough
Mark: you can continue to increase its robustness and something that the OEMs and other stake holders pay for
Michelle Wetterwald - Netellany [slides]
Michelle: I am making this on
behalf of ETSI, we are two partners TNO & Netally
... I wanted to share motivation for the project, requirements
and next steps
... basically you all know about IoT systems
... our project is about semantics, we have some definitions.
there are several standards for IoT, V2x, DATEX-II
... we need something to enable to interoperability, need to
have recurring aspects abstracted out
... it is now smart applications reference
... SAREF is an ontology connected with oneM2M
... it addresses a number of interests such as ehealth,
wearables, water
... we have two tasks to get the requirements from automotive.
we have collected use cases and identigy availabe existing data
models
... based on this input we are going to create a specification
of this ontology. it should be published in about a month,
approved by ETSI yesterday
... we are starting to create the actual ontology spec and
accompanying code. planned for release in June 2020
... what we have done so far is collect requirements. we have a
number of initiative like AIOTI, OneM2M, ETSI TC ITS, DATEX-II,
Sensories, and W3C Auto Ontology Group
... involvement in various around things like use case
platooning
... cooperative perception service, how to detect non-connected
vehicles
... we have a number of competancy questions [slide 9]
...
... we are going to include these into our ontology
[slide 10 on platooning competancy questions]
Michelle: we have a set of more than 100 of these questions, extending existing ontology made by SAREF and allowing/sharing from different domains
[slide 11 vehicle environment, vulnerable road users eg bikes...]
Michelle: the questions are not
exhaustive and hope to get it extended by those interested
after our initial release
... deliverable: ETSI TS 103. we will have a workshop in
Brussels later this year or early next and can share the
link
... please take a look at the technical report and provide
feedback
... we expect the ontology with technical spec to be published
in June 2020
KenV: I am somewhat concerned at a rather limited list of liaisons and wonder how others can get involved
Michelle: we would be very happy
to receive input and information. If you want to get access and
contribute you need to get involved with CEN to see how you can
liaison with ETSI
... I can help inquire if you get in contact with me
KenV: early access is important to increase awareness and get stake holders involved
Michelle: we are welcoming stakeholders and identify gaps
KenV: understand, trying to get involvement earlier on
Michelle: that is why I am sharing here and look forward to your presentation that you said you outlined like 20 SDO involved
George Pervical- OGC [slides]
George: We're excited to be in
part of this.
... The question we have is how do we characterize this
data.
... Turn things from device specific to device
independent.
... A couple are fundamental - space and time!
... Spacial Data on the Web overview
... Best Practice for Spacial Data
... Discussion on various best practices
... {Coordinate Reference Systems]
... Explains changes in characteristics of references
... As we get to decimeter accuracy, that level of accuracy
will become important for testing
... Sometimes it's lat/long others it's long/lat
... When you write your ontologies, worry about the order you
specify it
... [Time Ontology in OWL slide]
... {Allen Temporal Interval Algebra]
... [Semantic Sensor Network Ontology slide]
... Observatsions and Measurements slide
... These are used by NASA, intel events, Idian ocean typhone,
etc.
... We've gotten good at this abstraction
George: There are ways to state
where things are moving through space and time
... OGC has an interest in Connected and Autonomous
issues
... This diagram is from ETSI
... We may be thinkng about the temorality of some of these
things
... I don't know exactly what our scope is but we need to
collectively decide
Amir Sayegh - Geotab [slides]
Amir: you don't know
Google BigQuery to use it, just need simple logic to use the
data
… I give a talk about
why big data talks fail, this is about how it succeeds
… realtime streaming
apis and all that stuff
… this is a bit of a
commercial plugin to democratize our offering
… you can go into you
my geotab. I don't see people buying this as it is technology
that they will understand how to use
… what we have done is
tell out clients we can work with them as a bootstrapping
exercise, other lines of business data and mash them
together
… the glue that makes
it real is visualization. most of our data is geolocated in
nature
… one thing I was
telling my boss, when we send them csv slice it doesn't match
up to our potential based on data wealth
… we don't do that
anymore
Amir: Big Query visualizer tool
for loading large datasets gradually
... For example, analysis of fuel economy
... needed a quick way of identifying whether data is
clean
... using facet view
... AI as augmented intelligence
... open source real-time tracking app (planning to release on
github)
... Geohashes
... take the map, divide it into grids, can use public
standard, multiple resolutions
... snap-to-road is open issue
... Visualization on large screens at our Oakville
offices
... such as real-time weather map, with vehicle derived
data
... understand the biases in the data and anomalies
... one of our data scientists looked at the polar vortex last
year
... you could see the data disappearing as people stopped
driving
... data.geotab.com
... e.g. cell coverage dark spots, one fleet manager gave
instructions to avoid
... road impediments, Z axis of accelerometer
... hyperlocal weather network based on mobile sensors
... location analytics
... A weather station is millions of dollars, so a city has
few
... matching hyperlocal data with these stations' ground truth
can help both
... Intelligent Zoning
... we see drivers stop at one place, then another, as they
learn the drop-off location for an address
... so we can learn that zone information
... Chicago example, vehicle location model and patterns
... e.g school bus vs delivery vehicle
... Air quality measurement
... ID # of vehicles needed to do AQ measurement
... Intersection insights
... algorithms to identify intersection, signalized
intersection
... Aggregated Traffic Patterns
... e.g. for construction planning
DavidM are you involved with University of Toronto?
Amire: no but will be
GeorgeP: geohashes, squares, hexagons, S3. Various algorithms for grids
DickR: What connected vehicles is this coming from?
Amir: Geotab's vehicles
DickR: Zoning?
Amir: We can say, e.g., vehicles like you dropped off their deliveries here
Julián Rojas - Ghent University [slides]
Julian_Rojas: Ghent University,
Internet & Data Lab, imec
... Openplanner.team
... mobility as a service route planner
... combining road networks, rules and constraints
... public transport schedules
... shared mobility data, e.g. bikeshares and scooters
... transfer footpaths
... how you move among modalities
... e.g. how to transfer between bike and bus
... Challenges: how do we access; how do datasets interoperate;
how do we model it?
... tradtiionally, 2 options: data dump or API
<ted> i:Julian Route Planning on the Web
<ted> Slides
Julian_Rojas: We propose an
alternative, tiled or fragmented data
... interop
... linked data. RDF handles different data
serializations
... global identifiers
... linked data creates a framework for agreement
... data model depends strongly on route planning algorithm in
use
... Dijkstra-based, RAPTOR, CSA, Transfer Patterns each require
different data
... Linked Connections for public transport
... paginated
... Delaunay triangulation to identify transfer footpaths
... planner.js is js client-side library
... downloading fragments of data and triangulating
... because it's client-side, helps preserve privacy
... download data rather than sending endpoints
... data accessibility key to MAAS
... Standard semantics, should popular data models be
standardized?
Janina: can you plan a route with
accessibility considerations?
... e.g. with elevators not stairs?
Julian: yes, the data is out
there
... and tomorrow I'll talk about an applicaiton for
personalized route planning
... transit operators know which stops have ramps, etc.
Janina: also important to get real-time updates, for example elevator out of service
Julian: not aware of that data
@@: Could imagine a waze-like user input channel
Megan Katsumi - University of Toronto [Slides]
Megan: I'm from the University of Toronto within the Enterprise Integration Laboratory, our Director Mark Fox is here too
... knowldge based design, chain management...
... some of the early ontology work has taken place in our lab in the early 1990s
... defining organization roles, goals. work has been ongoing since then
... early work led to competancy questions
... more recently one result that has come out is ISO 21972
... this is a snapshot of the types of definitions you'll find there
... based on global city indicator ontologies
[slide 5]
scribe: these are using external, foundational (owl based) ontologies
... we're here today because of work in our city data model ISO WG11 Smart Cities
... goal is to standardize semantic interoperability across city services
JTC1
Megan: we have a three parts foundational ontologies, city level, services level
... time, location, observations etc are foundational for reuse. we use w3c owl time ontology for example as does everybody
... in city level we are reusing concept found in foundation level but conceptually apply at this level
Mark: foundation is true across all levels. at city level you may not use partof
Megan: services level is more an iterative development, adding a type of city service
... the starting point was iCity transportation planning, liaison with TC204 and looking at what other groups we can collaborate on
... we are looking to extend to include water and sanitation
... for transportation planning we are trying to figure out what the demand will be for the next 30 years
... not just road infrastructure but parking and all kinds of stuff
... problem is data silos, multitude of them and planning tools
[slide 10]
[[Proposed solution: ontology for transportation planning - part of iCity]]
[slide 12]
Megan: that is a quick review of what we are working on
[[what concepts are shared with other areas of transportation?
how can we work together to converge on these definitions?]]
Megan: we do not want our work to stand along but coordinate with ISO, OGC, W3C
Mark: the boxes in color are not notional but exist
... all of that is available on the web in OWL along with user manuals
... what we are doing via the ISO route is set the basis for the development of these standards. we are nearing the end of the project but now starting on the ISO route
... challenge to get stakeholders together now
... need to identify the areas for convergence and then mechanisms
... we will focus on different services beyond these
Joakim: I wasn't aware of this work, is it published?
Mark: as Megan said we are working with ESRI Canada and beginning work on translater between this and one from U Berkeley
@@?
Megan: we have some data fusion issues coming up
Mark: even the cities doing city transportation planning doesn't have access to eg household data - much isn't traditional city data
... we do handle EV and their charging stations
Josh: what are you using for your triple stores etc?
Megan: we are constrained with what we can use for free like alegra graph
Mark: you can attach metadata at the triple level
Ken Vaughn - Trevilon [Slides]
Ken: TC204 is a different committee than what Megan and Mark were describing (JTC1 WG11)
... pretty much every topic that has come up today can fall within our scope, not to discount work taking place elsewhere
... I am the convener of WG1 Architecture, one of the issues that came up is how we manage our data across our dozen WG with their own little silos
... we are seeing some national efforts to harmonize, making the problem larger
... rather than express our catalog of standards efforts done at various sdo and the use cases, focus instead on structure
... we need a top down approach to share data but to be honest I don't see that in our lifetimes...
... we need to give a top down and bottom up approach, pyramid paints a pretty picture
... let me discuss what we have done to date. how many are familiar with national ITS architecture, we are currently version 8.3 and initial goes back to 90s
... upgrade will include an information view, not the content of the view
... it will be populated over time
... every process we define we thing of data flows and a physical view. still abstract but takes those processes and assigns to physical objects
... it might be a roadside device like a traffic signal controller, vehicle itself
... some data flows will be internal to that device and we have largely ignored them
[its data model development approach slide]
Ken: information view is where things come together. we are currently trying to identify issues and gaps
... we rely on the community to learn about the problems that exist
[its reference architecture]
Ken: two important links ARC-IT http://arc-it.org and HARTS architecture http://htg7.org
... although HARTS is smaller it gets into more detail
... we are going to roll that back into ARC-IT
... a service package is roughly what you see in this diagram
Rudi: do you also reference information or code?
Ken: toolset to customize, I'll come back to that
... we have 1700 information transfers (triples)
[comm view]
[content of standard]
[info view]
DavidM: what group is responsible for level setting this back to analog?
...
<wseltzer> https://github.com/k-vaughn/its-reference-model
Ken: it is important these models are made available for wider industry adoption
... there is a fundamental problem in location/position between US and EU about what they consider the location of a vehicle center vs front center
... could be a problem
... need to figure out how to translate to logical standard
... I identified 24 standards bodies that should be part of these discussions
... we cannot delay current efforts but need to move to consistent models and a layered approach
[towards a consistent model]
Ken: we need some tools to guide us and help govern this process
... define data modeling framework
... what are the modeling rules? i need to be able to understand it quickly
... governance, where is it hosted, how do I submit commits and process
... need that model inter-sdo long term
Takuki_Kamiya: representing W3C WoT WG
... We have been working about 2 years
Takuki_Kamiya: Introducing WoT
... The IoT has connectivity standards, differing technologies
... then there's the challenge of creating applications that work across those technologies
... WoT is horizontal application layer for the web
... why the Web? user- and development-friendly
... applying those patterns to IoT
... https://xkcd.com/927/
... be the glue among existing frameworks
... Thing Description. collection of metadata
... interaction model, properties of actions
... Binding templates
... mapping to proptocols
... Scripting API
... cross-cutting concerns, security guidelines
... Published Candidate Recommendations. Architecture document, TD
... Abstract WoT Architecture
... Alignment with SOSA/SSN example
... Example TD
... We had a workshop in Berlin in June
... lots of ideas for phase 2, including industrial protocols, plug and play
... Issue: complex things. Is a vehicle a thing or a network of things?
... Expect to publish Proposed Recommendation after TPAC
... and propose recharter
Adnan Bekan - BMW [Slides ]
Adnan: Today BMW would like to present a positioning paper.
... Slide on Data-centric approach
... The problem we worked on was that if apps are designed to share then it could be problematic based on GDPR
... {Number of Connected Vehicles Increasing slide]
... [Web of Services slide]
... Context is important based on a high understanding fo the data.
... Where do you start, based on implemetation.
Adnan: We propose you start with a common goal
... At BMW how we do it is a pyramid starting with raw data that builds to information
... On top of information we build knowledge and then wisdom.
... We try to develop rules we can use (goes through 5 questions)
... What to standardize - we want to be a part of the IoT World
... In BMW we use Vehicle Signal Specification that started in GENIVI
... Everything starts with the vehicle in the physical world
... (builds hirarchy chart with branches}
... (describes branches and leafs)
... The vehicle taxonomy offers information.
... (reviews how VSS is working in W3C Auto WG)
... Within the BMW we have he Vehicle Shadow (contrasts old and new models)
... (stage 2 of Vehicle Shadow)
... Transformation is in backend but we intend to move that forward
... We ended up with two questions
... One possible answer proposed.
... Thank you.
Dick: If I understand this BMW will take their data to make it available to whom?
Adnan: We aren't use case driven. First we have to see can we make it available and then figure out who for.
Ted: A number of the OEM's we're talking to are interested - more on that tomorrow.
Steve Crumb - GENIVI [slides]
Steve: We started out getting Linux in autos and we did that a few years ago
... Now we're very keen on architectures.
...Steve: Slide 8
... Vehicle Cloud computing is the future
... Steve [Slide 9]
... Steve [Slide 10]
... Steve [Slide 11]
... This is a new initiative. We want to bring harmony to the current fragmented end-to-end picture
... Connected Car and Cloud architecture
... We want the industry to identify the building blocks that should be used together.
... [slide 13]
... [slide 14]
... This is open and investigative in nature
... no IP concerns, if you're willing to contribute your welcome
Steve: [slide 15]
... [slide 16]
... [slide 17]
... [slides 18 & 19]
Janina Sajka - Linux Foundation, Ted Guild - W3C https://www.w3.org/2019/Talks/tg-transport-a11y
Ted: We're here to explain a11y and why it's essential
Janina: Introduces background
... When we get the architetures correct we can change people's lives.
Janina: the closer data is to people, the greater the import of accessibility considerations
Ted: Web Accessibility Initiative is a great and passionate community at W3C
Janina: [slide 3]
Janina: APA is responsible for accessibility reviews
... including emerging technologies
... WCAG, Web Content Accessibility Guidelines
... ARIA is an overlay to map to accessibility APIs
... lots of work to define different accessibility needs
... and responses
... Ride-hailing: one thing we've done very well is ability to pay
... but identification of the vehicle could be better for blind riders
... we need a mechanism by which the rider can identify to the driver "I'm blind, here's my picture, look for me"
... a set of simple communicaitons would help
[Ted describes his introduction to W3C via Web Accessibility]
Janina: experiences are improving, but there are still gaps
... e.g. , in distance from vehicle to entrances
[slide 6]
[Factoring Accessibility into Data model design]
Janina: Bolted on solutions are worse than built-in
[slide 7: why introduce this topic now]
Ted: please follow up with us for questions
Janina: it really changes lives when you get accessibility right
David: I point you
toward JR Rizzo, and the AFL-CIO
... thanks for
this presentation
George: indoor
navigation and turn-by-turn
... for end-to-end
transportation
...
doorway-to-doorway
Josh: Are you aware of existing work on rideshare accessibility?
Janina: we could
start here
... brainstorming
user scenarios