Post on 22-May-2020
BIG Data for Automotive: What’s it worth?
Byron MasonSenior Lecturer Intelligent Systems
b.mason2@lboro.ac.uk
What is big data?
The volume of global data doubles every two years
@£0.001 Gb/month ~ £420 bn/mnth to store all the worlds data (lowest level of
redundancy)
y = 0.1728e1.2981x
0
10
20
30
40
2009 2011 2015 2020
Zetta
byte
s (1
x101
2G
b)
Global Data
1E-2
1E-1
1E+0
1E+1
1E+2
1E+3
1E+4
1E+5
1E+6
1980 1990 2000 2010 2020
Cos
t [U
SD]
Local data storage cost 1980 - 2020
megabyte gigabyte zettabyte
How much of big data is automotive?• High speed CANBus in
unconnected vehicles ~1Mb/s– Limited by storage <<100Mb
• Present day connected vehicles ~ >400 Mb/s– No local storage limitation
• 21m connected cars today generate ~ 8.4bn Mb/s
• 120 - 250m connected cars by 2020
• Autonomous vehicles ~ 40.5 Gb/s
How much of big data is automotive?
• Assuming the average person spends ~1hr in a car and has 1.265 cars.
• With 1.215bn cars in use globally (2015).
• Assuming 16% are connected, that is 2.80 x 1011
Gb/day or 1.02 x 1014 Gb/yr.• @$0.005 /Gb/mnth for
enterprise grade storage, the cost is £417bn/year.
OEM (Global Market share)
Storage cost / yr by market share (2016)
Cost relative to revenue (2017)
Toyota (9.2%) 38.40bn 17.80%
Volkswagen (7.2%)
30.02bn 14.14%
Ford (6.5%) 27.11bn 21.20%
Honda (5.5%) 22.94bn 21.30 %
Some estimates suggest that the value of data generated by a car over its life will be equal to its
retail price
DATA ≠ INFORMATIONData is a set of values related to variable(s)Information is the resolution of uncertainty
DATA + ANALYSIS = INFORMATION
The business opportunityCar data monetisation expected to be worth 450 – 750bn USD dollars by 2030, compared with ~ 4tn USD vehicle sales in the same period (18.8%)
Source: McKinsey
The powertrains opportunity: todayOn road diagnosis using vehicle telemetry data
– Reduced (50%) warranty costs– Improved fault diagnostics for
expedited repair– Increase commercial vehicle
availability by 10%
Predictive maintenance using vehicle telemetry data
– Repair before failure– Quicker root cause analysis
Improved development process– Preventing recurrence of
manufacturing issues– Engineer tacit knowledge capture
and congnitive enquiry
Extracting the value: fleet based development
• Component engineering choices have fleet consequences
• Engineering at fleet level considering;– Commodity prices– Labour prices– Target market
• To minimise– Fleet fuel consumption– Cost– Complexity
• Data is readily available• Informed decision making at
engineer’s level | Augmenting Intelligence
Analytical optimisation | Physical and Machine Models | Cognitive compute |
Extracting the value: Calibration customisation and adaption• Data sources;
– Powertrain/vehicle sensors– Pollution information– Traffic information– Weather information– Day / Time (is it Christmas?)– Terrain maps– Driver style– Cycle classification
• Calibration selection (5 – 7%) fuel improvement for driver/cycle/terrain
• Pay for performance ‘updates’ Learning controller / over the air updating
BEV Controls Customisation• Energy management is critical and tends to be
conservative to extend range.• Use of classifiers to determine probability associated of;
– Drive cycle i.e. driving to work, shopping, Sunday drive.– Vehicle loading along the route– Stop off points
• Use prior knowledge– Terrain– Weather (wind speed and direction)– Google ‘business’ indicators
• To inform controls decisions for;– ‘Sporty’ response– Range maximisation– Preheat (battery and cabin)
• And present marketing opportunities;– Spouses birthday present?– Coffee before work?– Lunch on a long drive
"Information is the oil of the 21st century, and analytics is
the combustion engine”
(Peter Sondergaard, Senior Vice President, Gartner)
Project partners
References
• http://www.mkomo.com/cost-per-gigabyte• https://setis.ec.europa.eu/system/files/Driving_and_parking_patter
ns_of_European_car_drivers-a_mobility_survey.pdf• https://www.greencarreports.com/news/1093560_1-2-billion-
vehicles-on-worlds-roads-now-2-billion-by-2035-report