1

Confidential

5/21/17
AUTONOMOUS
VEHICLES:
FORWARD GEAR
Advantages of
Autonomous Vehicles
Efficiency through Automation
Learning advantage:
Humans may repeat
the same mistakes
as millions before
them, the
automated vehicles
will draw data and
experience from
thousands of other
vehicles on the road.

By pulling in data
from complementary
sensors from other
nearby vehicles
Automation will
(vehicle-to-vehicle or
reduce the role of V2V) and
human choices and infrastructure such
behaviour in driving. as traffic lights and
road signs (vehicle-
to-infrastructure or
V2I), the
performance of such
autonomous systems
Reduction in accidents

1.2 13
50 94 2
5 million %
Thousan
million
d
Deaths Each Suffer from Crashes tied Deaths each
Year globally fatal injuries to a human year in
due to road due to road choice or Indians road
accidents. accidents. error.* accidents.

*drunk driving, excessive speed, distracted driving

etc.
Inclusivity

People Unconducive
with Economic
Disability Conditions etc.

Elderly

Persona
Safe
l
Benefits through pooling..

Common routes will

Further, an Net result will be
be deciphered and
autonomous vehicle fewer cars on the
deployed
will never be road and freeing up
accordingly
parked idly. It will of physical space
whereby multiple
be in continuous for utilization while
users will have to
deployment. was earlier used for
use only one such
parking.
vehicle at a time.
Market for AVs

Boston Consultancy Group estimates that 12 million AVs sold globally

with a market for partial and full autonomous vehicles at $42 billion by
2025, and 18 million vehicles and market of $77 billion by 2035. [1]
McKinsey estimates that on-demand mobility services and data-driven
services could create up to ~$1.5 trillion (or 30 percent more) in
additional revenue potential in 2030, compared to ~$5.2 trillion from
traditional car sales and aftermarket products/services (up from ~$3.5
trillion in 2015). [4]
 AV Long-Term Growth Projections

Fully Autonomous Vehicles Partially Autonomous Vehicles2

(In Billion USD. Source: Boston Consultancy Group, AV Adoption Study)

Forces at play
Forces

Regulation Global, National, State and Local level legislations and regulations
Taxation

Social Attitudes Perception, trust in machines

Progress towards shared economy; longstanding notions of vehicle ownsership

Technology Learnings from early experiments

Innovation or technological breakthroughs
Cost and reliability
Privacy & Cyber-security and Communication standards and protocols
Security Protection of personal identification information
Laws relating to data privacy and protection
Capital Availability of investment capital

Key Stakeholders Potential changes to current employment models, including dislocation effects, costs and change
management & Future employment growth opportunities (nature and size).
Stakeholder reactions and next steps. (eg. Workers, unions, dealers, employers, government etc)
Technology Used and limitations

Because much of the development of driverless cars is proprietary, the

exact capabilities and their extent are unknown.
Technology Overview Limitations
LIDAR Light Detection and Ranging- Noise removal, interpolation to
Senses and evaluates the fixed point spacings.
distance to an object by using Triangulation issues.
light.
GPS Satellite navigation system The accuracy is not very high
providing location and time. (+/- 10 metres). Lose signal
easily.
DGPS Differential GPS - enhanced form Loses signal 1m of accuracy
of GPS, highly accurate (+/- every 150km. Shadowing from
10cm). buildings, underpasses, foliage
etc causes signal loss.
RTK Real Time Kinematic satellite
navigation, data from GNSSs viz-
a-viz a reference station.
Cameras Super high definition cameras, HD capture requires a lot of data,
capable of capturing colour highly computational extensive
unlike others. High availability and algorithmically complex.
and low cost.
Digital Maps Collection of real world Limited areas mapped, Require
infrastructure, road, transit, extremely high level of detail
traffic data in virtual image. and frequent updates.
Current Issues

Vehicles with Human Error

Semi- At present due Early
Autonomous to regulatory assessments
and Fully and technical have shown
Autonomous reasons the accidents Inadequate
Capabilities Autonomous regarding Infrastructure
are already Vehicles may Autonomous
deployed on only operate in Vehicles occur
roads, in the presence due to due to
certain of a human the following Malfunctioning
jurisdiction; driver; reasons: Software
Contd

Additionally, the ability to

Affixing liability in Accidents make decisions based on
involving Autonomous ethical considerations cannot
Vehicles be programmed into
Autonomous Vehicles;
Human Error

Lack of a uniform
approach by all the
different parties
Most of the developing such
existing systems. There are
infrastructure, varied
Human on-road being designed for implementations of
behaviour is humans, would the technology,
unpredictable. necessitate regular potentially creating
In the current human major regulatory
trials, most of the interventions and and safety issues.
accidents took inputs in the
Because the place due to human autonomous
existing fleet of errors (usually by process, to ensure
non-autonomous the other vehicle). proper
cars cannot be More caution will performance. [2]
replaced at once, have to be
the self-driven cars exercised by self-
shall have to driven vehicles in
operate alongside such situations.
human-driven cars
for the foreseeable
future.
 Inadequate Infrastruture
always

Existing infrastructure designed for human driving.

not always
conditions not
conducive
conducive
Road conditions

Weather conditions such as snow, rain or hail greatly reduce

effectiveness of self-driven cars.
Road

Heavily dependent on the state of mapping in that specific

location, which may not always be accurate or up-to-date.
Malfunctioning Software

It is crucial that the

Operating System in
Autonomous Vehicles
be free from errors.
However at present,
no software exists
which is 100% free of
glitches or bugs
Possibility of third-party
modification of on-board software

Even most
secure The
The efficacy of
software can modifications
such software
be usually so done will
is also not
hacked or not be within
completely
broken into the regulatory
reliable at all
(such as framework of
times.
jailbreaking review.
of iOS).
Possibility of unauthorised remote
access/hacking of on-ward software

Critical vehicle functions

The hacking can also be
like speed control,
unauthorised, by
braking, door locks etc at
malicious parties.
risk of being hacked

Real life incident of Tesla

Model S car having its
critical functions being
taken over by Chinese
hackers from 12 miles
away
 Liability Related Issues

Since autonomous
vehicles are
Should stricter standards to be applied
perceived to be in such cases?
safer than human
driving, the
question remains
that should the
allowed margin of What will be considered as the
error be higher for standard of reasonable care in such
autonomous cases?
vehicles than in
the case of
humans.
Current Framework

Strict Liability
Liabil
ity Negligence Law.
can Breach of Warranty
take
vario Consumer Protection laws
us
forms
 In case of
autonomous
cars, there exists
an issue in
attributing
liability (of any
nature) as two
models of
development
exist:
The same entity One entity develops software
develops the (such as Google, Uber etc) to
be used on a vehicle
onboard software manufactured by another
and manufactures entity (such as Ford, Volvo
etc). This software in installed
the vehicle. (Such either before it is sold to the
as Tesla, Mercedes, final customer or as an
aftermarket addition for non-
BMW etc). autonomous cars.
 Determining the exact cause
or source of an untoward Determining liability when
incident, based upon which the on-board software has
the software developer or the been modified, either
hardware manufacture would intentionally or by
be held liable, would prove to unauthorised hacking, which
be difficult, as well as an then leads to injury would
unnecessary burden on the create further issues.
injured party.
Depending on the degree of
automation, as well as the
situation at the time of injury,
liability would have to be
Volvo, in Oct 2015, declared
determined to be placed on
that it would accept full
either the human (driver or
liability for whenever its cars
otherwise) or the on-board
are in autonomous mode.
software. In case of
coordination between the
two, attribution will become
further difficult.
 Prominent Current Players

Development of AVs across the globe.

USA Google, Tesla, Uber, Ford, General Motors etc.
Germany Mercedes-Benz, BMW, Audi, Bosch etc.
France Citroen, Renault, Peugeot etc.
Sweden Volvo, Autoliv etc.
Japan Nissan, Toyota, Honda etc.
China Baidu, Chongqing Changan etc.
Singapore NuTonomy, Delphi etc.
(Bold indicates on-road trials of their full autonomous vehicles.)
Currently On-Road

Google
Started trials in 2012 in California. 2,230,275 miles driven in autonomous mode by Oct 2016.
Based on Google Monthly Reports, 16 crashes reported so far, but only 2 due to the on-board software. Rest attributed to
human errors.
Tesla
Allows a form of semi-automation on its existing mass market vehicles called Auto-Pilot, allowing functions such as
automatic lane changing, parallel parking, remote retrieval from parking space etc.
Made provisions for full SAE level 5 automation from 2017.
On May 7th 2016, a driver using Auto-Pilot, Joshua Brown, died in a collision with a truck due to an issue with the on-board
software. (The Guardian, 1st July 2016).
Uber
Started on road trials in Pittsburgh in Sept 2016 .
14 autonomous vehicles, with a backup driver and engineer, ferrying passengers around.
Regulation of
Autonomous Vehicles
Degrees of Automation
SAE International
released classification
of levels of automation
of on-road vehicles.
(Doc No. J3016).
This classification was
also adopted by the
Dept. of Transport
(USA) in their 2016
policy regarding
automated vehicles.
Basic premise behind
these levels: who
does what, when
(DoT, 2016).

(Source: SAE International, Document

Global Regulatory Roundup

USA:
Has the first-mover advantage, as it was the first nation to allow autonomous vehicles to ply on
roads;
The current federal rules allow for AVs if they abide by FMVSS and have a conventional design.
Nevada was first state to allow Autonomous vehicles on road and since then more and more
States have allowed on-road trials of autonomous vehicles
EU
Several EU countries have allowed on-road testing of Autonomous Vehicles after 2013
European Transport Safety Council is working towards is trying to develop a scheme for
providing approval to autonomous vehicles
Such framework would involve updating of driving license regulations keeping in mind unique
needs of automated driving systems and upgrading the infrastructure
 CHINA
The Auto Industry Regulator in China is in the process of drafting rules for testing of
Autonomous Vehicles
Till the rules are complete it has banned the testing of Autonomous Vehicles on highways
JAPAN
As per the guidelines published by the National Police Agency, on road testing of
Autonomous Vehicles in Japan can only take place in the presence of a driver;
Fully Autonomous Vehicles will have to have black boxes to record vital data
A panel of experts has also be constituted to suggest the required legislative changes
SINGAPORE
SAVI or Singapore AV Initiative collaboration between Land Transport Authority (LTA)
(regulatory aspect) and Agency for Science Technology and Research (A*STAR)
(development of technology).
SAVI identified one area of Singapore as testing ground.
LTA collaborating with Delphi and NuTonomy, under SAVI, for AV trials.
A new Centre of Excellence for Testing and Research of Autonomous Vehicles-NTU
(CETRAN), a collaboration between LTA, NTU and JTC, for development (simulations,
cybersecurity etc) of AVs.
Testing parameters

Usually, regulations mandate that vehicles be test driven for a certain amount of time/distance to
establish their roadworthiness, safety and emission credentials etc. prior to public road use.
However, due to the sheer number of variables and factors involved in fully autonomous vehicles,
AVs would have to be driven hundreds of millions of miles and, under some scenarios, hundreds
of billions of miles to create enough data to clearly demonstrate their safety and reliability.
Under even the most-aggressive test driving assumptions, it would take existing fleets of
autonomous vehicles tens and even hundreds of years to log sufficient miles to adequately assess
the safety of the vehicles when compared to human-driven vehicles, according to the analysis.
Researchers say the findings suggest that in order to advance autonomous vehicles into daily
use, alternative testing methods must be developed to supplement on-the-road testing.
Alternative methods might include accelerated testing, virtual testing and simulators,
mathematical modeling, scenario testing and pilot studies. [1]
Toyota has set a target of 8.8 billion miles of testing before its AVs reach the customers. [2]
Federal Automated Vehicle Policy 2016

In US the Federal and State roles for motor vehicle operation are clearly demarcated.
Federal authorities are responsible for regulating motor vehicles and equipment
(FMVSS, Investigations of motor defects and recalls etc) while State authorities
regulate the human driver and most other aspects of motor vehicle operations
(Licensing of Humans and vehicles, traffic laws, insurance and liability etc).
If the new vehicles are compliant with the existing Federal Motor Vehicle Safety
Standards (FMVSS) and maintain a conventional design, then no specific federal
legal barrier on sale of HAVs.
USA, via Dept of Transportation, released this policy in Sept 2016. They are agency
guidelines more than rulemaking.
Primarily addresses High Automated Vehicles (HAVs) [Based on SAE International
classification system as A3 to A5 levels of automated driving].
Promotes data recording and sharing capabilities, while ensuring privacy,
to allow government, industry and the public to increase their learning and
understanding of the technological advancements and challenges. Best
practices to be shared and implemented.
Encourages all specific HAV systems to clearly the define the Operational
Design Domain (ODD) and their corresponding SAE levels, as well the
conditions in which what function is intended to operate w.r.t. roadway
types, geographical location, weather conditions, speed range etc. This is
ensure the requisite safety Object and Event Detection and Response
(OEDR) systems are in place in case of failures.
Includes a Model State Policy to be adopted by the states for a uniform
and consistent regulatory environment.
Insurance and liability issues are left to the States to decide for
themselves.
Rounding up, the policy is supportive of the development and adoption of
Automated Vehicles in the USA, while being mindful of the requisite safety
and security considerations for the stakeholders involved.
(Source: Federal Automated Vehicle
Regulatory Challenges in India

Regulatory framework in India is suitably malleable to allow for

changes.
Policy makers rarely take advantage of this flexibility.
Instead of setting standards, we rely on standards set abroad.
Current Procedure for Introducing Prototype Vehicles on Road

Vehicle has to fall in one of the five categories: Light Passenger

Vehicles, Commercial Vehicles, Tractors, Two Wheelers and Three
Wheelers.
Type Approval Extensive laboratory testing for safety critical
components and assembly by iCAT/ARAI.
Testing is based on notifications issued by Ministry of Road Transport
and Highways.
Notifications tend to be very specific stating not only end result, but
also components to be used.
Only exception is for less than 25cc vehicles.
Suggestions for Future Regulations

Any program for assessing, testing and approving autonomous vehicles be accessible across all segments of the
industry.
Such programme must be devised by taking relevant inputs from all stake holders including- Government (Central
and State); Auto OEM; Regulatory Bodies; Academia and Citizens Forum & Telecom (5G) and Smart City
(Intelligent Transportation System or ITS).
Basic capabilities of AVs may be tested by conventional tests.
Once they clear conventional tests, they may be deployed for on-road testing with additional safeguards, such as:
Limiting areas in which they operate.
Always having a human driver present, even if vehicle is in autonomous mode.
Higher qualifications for drivers to operate AVs.
Lower speed limit.
Greater distance to be maintained between AV and other vehicles.
Fail-safe mechanism with manual override in case of emergencies.
Limiting number of AVs on road to ensure proper monitoring.
Data garnered from such testing stored at a centralised server and available to public at large.
Infrastructural Challenges

Creating infrastructure that can be easily read by AVs.

Clearly Demarcated Roads Road signs that will be visible and apprehended by the AVs in all
weather conditions (such as heavy snow, fog, rains and glare from the sun etc).
Sensors along lanes Intersections.
Traffic Signals Should (ideally) be able to communicate to the AV through V2I technologies, and
V2V data from other vehicles on the road, as well as provide the option for data collection and
aggregation through ITS technologies so AVs have live updates regarding the entire surrounding
traffic grids in advance.
Stricter traffic and pedestrian rules AVs are unable to distinguish between the alertness levels of
pedestrians (such as if theyre distracted by their phones etc), and unable to respond accordingly.
Specific areas and lanes - Need to be developed where only AVs will operate, to minimize conflict
with human driven vehicles.
Data sharing, storing and analysis between various entities (both private bodies and governmental
authorities). Will require a form of standardization of certain key data points.
Recent Developments

Dr. Roshy John, Practice Head of TCS built a autonomous version of Tata Nano in Feb
2016. This technology can be used in any vehicle, installed in an hour, according to his
team.
Novus Drive, developed by The Hi-Tech Robotic Systemz Ltd of Gurgaon, an indigenous
driverless shuttle, demonstrated at 2016 Auto Expo.
DMRC is incorporating driverless trains, starting with two routes, in its phase-III
expansion, to be in place by early 2017.
Tata Elxsi is part of autonomous car programme and has developed a new concept that
allows for automated vehicle parking, effectively communicating with parking
gateways/intelligent servers and collecting information about available parking slots. [1]
Mahindra Reva, the companys electric car subsisdy, has already initiated R&D in
Bengaluru and hopes to begin testing these cars once the requisite approvals are
obtained. [2]
Thank you