GTC 2017 AUTOMATED TRUCK DRIVING AND PLATOONING …on-demand.gputechconf.com/gtc/2017/presentation/s...2015 2008-2011 Intersafe II 2008 -2012 SIM-TD 2010-2013 2011-2016 2011-2013 Drive

© fka 2016 · All rights reserved2017/05/09Slide No. 1·

San Jose, 9th May 2017

Dipl.-Ing. (FH) Devid Will, M.Sc., Dipl.-Ing. Jens Kotte

AUTOMATED TRUCK DRIVING AND PLATOONING WITH DRIVE PX 2

GTC 2017

Forschungsgesellschaft Kraftfahrwesen mbH Aachen


MotivationTruck related efficiency topics (focus EU)

Source: dieterblasl

En

erg

y E

ffic

ien

cy

Source: t-online.de

Tra

ffic

Eff

icie

nc

y

Source: stuttgarter-zeitung.de

Sta

ff E

ffic

ien

cy

29%

26%9%

9%

27%

Staff

Fuel

Toll

Administration

Other

Source: www.bgl-ev.de

Average Portions of Costs in 2013 (Germany)

Co

st

Eff

icie

nc

y


MotivationTruck related safety topics

Source: remseck.freiewaehler.de

20%

17%

12%10%

41%

Distance Turning/Switching Priority in Traffic

Speeding Others

Accident causes in 2014 (Germany)

→ Too small gaps are the main reason for truck accidents in Germany!


Basic truck ADAS informationToday‘s available truck DAS

Source: trucknews.com

→ Introduction of new truck Advanced Driver Assistance Systems (ADAS) only if mandatory or a clear business case for

fleet owners available!

Source: DAF

Source: DAF

ACC

Source: Mercedes Benz

PPC

AEB LDW


MotivationTruck related driving pleasure topics

Source: express.deSource: lasiportal.de


Technology transfer passenger vehicle to truckExamples for challenges

Source:beschriftungscenter-fuechsl.deSource:fahrzeugbilkder.de

Configuration Variations Large Pitch Angle

Source:mercedes-benz.co.uk

Dimension Variations


informed assisted partly automated highly automatedfully

automateddriverless

Platooning…

Robust & secure

connectivity … C-ACC Platooning

Google

Car

without

controls

…„Classic

Car“

no

connectivity…ACC

Highway Assist

(ACC+LKA)

autonomous

highly

automated

driving

Highway Pilot

e.g.

traffic information,

eCall

e.g.

radio based

danger warning

Highway Chauffeur ……state-of-the-art

connectivity

Level 0 Level 1 Level 2 Level 3 Level 4 Level 5

System

L e

v e

l

o f

C

o n

ne

c t

iv

it

y

L e v e l o f A u t o m a t i o nHuman

Connectivity & automation Levels of automation and levels of connectivity


Highway Assist

(ACC+LKA)

Highway PilotHighway Chauffeur


automateddriverless

Platooning…

Robust & secure

connectivity … C-ACC Platooning

Google

Car

without

controls

…„Classic

Car“

no

connectivity…ACC

autonomous

highly

automated

driving

Highway Pilot

e.g.


eCall

e.g.

radio based

danger warning


connectivity


System

L e

v e

l

o f

C

o n

ne

c t

iv

it

y



Driver in

the Loop

Driver has to be able to take

over control within a certain

time (e.g. construction zone)

System can handle

highway specific

situations. In case of an

unknown situation it goes

into a secure state.


C-ACC PlatooningPlatooning PlatooningPlatooning


automateddriverless

…

Robust & secure

connectivity …

Google

Car

without

controls

…„Classic

Car“

no

connectivity…ACC

Highway Assist

(ACC+LKA)

autonomous

highly

automated

driving

Highway Pilot

e.g.


eCall

e.g.

radio based

danger warning


connectivity


System

L e

v e

l

o f

C

o n

ne

c t

iv

it

y



• Often called Platooning

• System only takes over

longitudinal control

• Gaps > 15 m

• System takes over

longitudinal & lateral control

• Gaps < 15 m

• Driver has to take over control

in certain situations (e.g.

construction zone)

• System takes over

longitudinal &

lateral control

• Gaps < 15 m

• System is the

ultimate fallback

level


PlatooningOverview & goals of platooning system

• Improvement of safety due to longitudinal and lateral

guidance

• Relieve and support for professional drivers

• Improved road space

• Optimization of traffic flow

• Reduction of fuel consumption due to slipstream driving

Goals

Source: man-truckers-world.de


PlatooningDifferent types of platooning

Gap between vehicles

Longitudinal Control Lateral & Longitudinal Control

▪ Also called “Cooperative ACC”

▪ Driver in the following truck still

has to steer

▪ Level 1 system

▪ Following trucks follow first

truck automatically

▪ Following truck driver does not

have to steer, brake or

accelerate


MFG

EFAS

2001 - 2005 INVENT

2006 - 2010 AKTIV 2012-2016 UR:BAN

interactIVe

eCoMove

eValueMOTIV

FP5

1992-1994

TAP (FP4)

1994 - 1998

1992 - 1994

1985 1990 1995 2000

PROMETHEUS (EUREKA)

IST (FP5)

1998 - 2002

FP6

DRIVE II

FP7

2007 - 2014

DRIVE I1989-1991

PROTECTOR

CHAMELEON

PROMOTE CHAUFFEUR II

2002-2005

2000-2003

2000-2003

PREVENT

PROMOTE CHAUFFEUR I

ADASE II2001-2004

RESPONSE 22001-2004

KONVOI

1996-2000

European

German

AIDER 2008-2011

euroFOT2008-2012

TeleFOT2008-2012

2015

Intersafe II2008-2011

SIM-TD2008 - 2012

2010-2013

2011-2016

Drive C2x2011-2013

Pre-Drive C2x2008-2010

SARTRE

ecoDriver

Assess2009-2012

HAVE-it

2004-2008

Transport Telematics (FP3)

2020

2014 - 2020

AdaptIVe

VRA-Net

AutoNet2030

iGame

Companion

2002 - 2006

2005

1987-1995 1996-1998

2000-2003

2003/4

2005 - 2009

2001/2

2010-2013

2009-2012

2008-2011

2014-2017

2013-2016

2013-2016

2013-2016

2013-2016

2010

© ika/fka

SCOUT2016-2018

CARTRE2016-2018

Pegasus

ART-042016-2019

ART-02

2017-2020

ART-03

2018-2021

2015-2018 Ko-HAF

2016-2019

Enable-S32016-2019

Platooning

Status quo on automated drivingA highly relevant research topic for many years


Platooning ProjectsKONVOI

Goal

Practical usage of truck platoons in road freight transport for verification of prognosticated effects (economy of road space,

reduction of fuel consumption, …)

Project overview

• Duration: 2005 – 2009

• Funded by the German Federal Ministry for Economic Affairs and Energy

• Project Partner: RWTH Aachen University, BAST, MAN, WABCO



• Electronic coupling of two or more trucks.

• Target distance between vehicles is 10 meters.

• The KONVOI-system includes lateral and longitudinal guidance.

• The leading vehicle will be driven manually.

• The following vehicle will be automatically driven in longitudinal and lateral direction.

• The system is monitored by a truck driver at all times.

• The KONVOI-System can be overruled by the driver at any time.

Functionality of the system



Overview test truck equipment

Brake booster

Distance sensors

Image processing

GNSS position determination

In-Car PCs & microcontroller

Digital map

Vehicle-to-vehicle communication

Steering torque and angle interface

dSpace AutoBox



Test runs on public highways in 2009

→ First operation of platoons in real traffic worldwide!


Platooning ProjectsEuropean Platooning Challenge 2016


Automated DrivingFrom Sensing to Trajectory Planning

Automated Driving can typically be divided into

▪ 1. Perception: Digitalization of environment – object detection, freespace detection, characterization

▪ 2. Localization: Mapping of objects into local coordinate system

▪ 3. Motion Planning: Behavior generation and trajectory planning

▪ 4. Actuation: Trajectory tracking and vehicle dynamics controller

Perception &

Localization

Behavior

Generation

Trajectory

PlanningActuationSensing


Motion Planning for Assisted and Highly Automated Cars/TrucksTrajectory Planning

Simplification of the trajectory planning problem

▪ Find a suitable control function, i.e. a steering and speed profile, which

▪ guides the vehicle in a safe and comfortable manner through the environment

▪ respecting vehicle’s and environmental constraints, e.g. max steering angle, collision avoidance,

vehicle tilting limit, road friction

Control Function Vehicle Model & Rating Drivable Trajectory

Obstacle

Ego-Vehicle

State Trajectory


Cooperation Between NVIDIA and fkaHistory and Future Plans

History

▪ First contacts in late 2014

▪ Loose conversations and meeting in Santa Clara, CA

▪ fka presented its motion planner @ELIV 2015, Baden Baden

▪ Contact to NVIDIA

▪ Further exchange on phone, physical meeting @fka

▪ fka and Nvidia become partners, fka gets full access to Drive PX

▪ fka is invited to join NVIDIA at its booth @CES 2016

Future plans

▪ Widen partnership in industrial projects

▪ Possibly integrate motion planner to

DriveWorks software stack








Perception &

Localization

Behavior

Generation

Trajectory



Neural nets for lane marking recognition (based on AlexNet)Comparison of three different nets

Complex Normal Thin

→ All nets very computational expensive and not executable in real-time on Drive PX 2

→ Using NVIDIA’s LaneNet on Drive PX 2


NVIDIA LaneNetRunning on Drive PX 2 with parameter extraction


Control parameter derived from LaneNetControl input for lane following control








Perception &

Localization

Behavior

Generation

Trajectory



Phone

Fax

Email

Internet www.fka.de

fka Forschungsgesellschaft Kraftfahrwesen mbH Aachen

Steinbachstr. 7

52074 Aachen

Germany

Contact

Dipl.-Ing. (FH) Devid Will, M.Sc.

+49 241 80 25676

+49 241 8861 110

[email protected]

Documents

GTC 2017 AUTOMATED TRUCK DRIVING AND PLATOONING …on-demand.gputechconf.com/gtc/2017/presentation/s...2015 2008-2011 Intersafe II 2008 -2012 SIM-TD 2010-2013 2011-2016 2011-2013 Drive