Cha halleng nges in n archi hitecting ng fully y autom omated - - PowerPoint PPT Presentation

▶
cha halleng nges in n archi hitecting ng fully y autom
SMART_READER_LITE
LIVE PREVIEW

Cha halleng nges in n archi hitecting ng fully y autom omated - - PowerPoint PPT Presentation

Oct 04, 2023 280 likes •690 views

Cha halleng nges in n archi hitecting ng fully y autom omated drivi ving ng; on Heav avy Commercial al Veh ehicles es with a h an n empha hasis on 1 Naveen Mohan et al. ; WASA 2016 To o the he b best of of ou our know

slide-1
SLIDE 1

Cha halleng nges in n archi hitecting ng fully y autom

  • mated drivi

ving ng;

with a h an n empha hasis on

  • n Heav

avy Commercial al Veh ehicles es

Naveen Mohan et al. ; WASA 2016

1

slide-2
SLIDE 2

Mai ain contrib ibutio ions

To

  • the

he b best of

  • f ou
  • ur know

nowledge, this is paper is is uniq ique in in

➔Full r ll range of possib

ibil ilit itie ies o

  • f

int ntegrating ng i int ntelligenc nce to

  • an

n autom

  • mot
  • tive p

platfor

  • rm

➔ Di Disc scussi ssion across ss a a broad ad sp spectrum of

  • f asp

spects w.r .r.t. autonom

  • nomy bot
  • th func

nctiona nal an and extra-func nctiona nal

Autonom

  • nomy mi

mindma map

2

slide-3
SLIDE 3

: Case s study planned : KTH Res esea earch c concep ept veh ehicle, e, Scan ania t a truck

3

slide-4
SLIDE 4

Results: Case 3 vs Case 4

1 2 3 4 5

Higher Platform Reuse Lower accidental Complexity(on reuse) Lower Variability(across platform) Lower Development Cost Upfront Lower Development Cost over time Higher Reliability/Availabilit y Reduced need for Diagnostics to ensure safety Higher Security Ease of Verifcation

  • f Modified Pf

functionality Ease of Verification

  • f ADI functionality

Lower Information flow needed and infrastructure

Case 3 Case 4

4

slide-5
SLIDE 5

www. www.kth.s .se/ it itm/ autonom

  • nomymind

ndmap

Join and collaborate. KTH: Mechatronics ITRL IC

5

slide-6
SLIDE 6

Bio:

  • : Naveen M

n Moha

  • han

➔ Bachelor’s in Computer Science and Engineering (2009 009) ➔ 1 year; Defence Industry; Communication, Networks ➔ Master’s in Networks and Distributed Systems (2 (2012) ) Chalmers, Gothenburg

➔ 3 ye years; Automotive ve Indust stry; VCC; SW/ System responsible Hybrid, el drive ➔ PhD hD studies es a at M Mec echa hatroni nics KTH ( (Started ed end end of 2 2015); The he ARCHER ER proje ject: Vinno nnova fund nded ed

6

slide-7
SLIDE 7

By Veronica538 (Own work) [CC BY-SA 3.0 (http://creativecommons.org/licenses/by-sa/3.0) or GFDL (http://www.gnu.org/copyleft/fdl.html)], via Wikimedia Commons

What we are trying to do?

7

ADI = Autonomous Driving Intelligence

slide-8
SLIDE 8

No Reuse Full Reuse

Case 1 Case 2 Case 3 Case 4 Case 5

Preserve legacy Optimize for functionality

9

slide-9
SLIDE 9

Key Messages

➔ The e role o e of l leg egacy ➔“intel elligen ence( e(ADI) in integratio ion” ➔The he drive ver ha has to go! ➔Sa Safety need eeds to b be e proven en ➔Prototyp ype vs vs Pr Product

10

slide-10
SLIDE 10

Ou Outline

➔ Ab About the e Author(s (s) ) and nd the he projec ect ➔ Bac Background

➔ Complexity an and Legac acy ➔ Advan antag ages of autonom

  • nomou
  • us HCVs vs

passen enger er veh ehicles es

➔ Case ses: s: ADI DI i integration. ➔ Conc

  • nclusions
  • ns, future w

wor

  • rk a

and nd questions

  • ns

11

slide-11
SLIDE 11

By Andy Dingley (Own work) [CC BY 3.0 (http://creativecommons.org/licenses/by/3.0)], via Wikimedia Commons

The he rol

  • le of legacy in

n automot

  • tive

sy syst stems s desi sign

➔ Accident ental vs es essent ential complexit ity ➔ Legac acy as as a s a source of accident ental c complex exity ➔ Modularit ity im implie ies that n no vehi ehicle i e is optimized ed i in t n ter erms

  • f functionality.

y. ➔ Ther here e are e differ erent ent ways to achi hiev eve e the he same e functio ionalit ity ➔ The he impact of l leg egacy

12

slide-12
SLIDE 12

?

(Exagger erated ed ex example) e) Design c n cons

  • nsiderations
  • ns

CC BY-SA 3.0, https://commons.wikimedia.org/w/index.php?curid=342457 By Andy Dingley (Own work) [CC BY 3.0 (http://creativecommons.org/licenses/by/3.0)] via Wikimedia Commons

13

slide-13
SLIDE 13

One Com

  • mmon
  • n Electrical Platfor
  • rm

1.2

14

slide-14
SLIDE 14

1500 00 logical al nodes es

A Scania production vehicle from 2013

15

slide-15
SLIDE 15

14000 000 conne nnections ns

A Scania production vehicle from 2013

16

slide-16
SLIDE 16

Dealin ing w wit ith complexi xity

➔We e compensate.. .. ➔Arch chitect ctur ural mechanism sms. s. ➔Pl Platforms ➔Process m ss measu sures ➔Stan andar ards ➔Stan andar ardizat ation

17

slide-17
SLIDE 17

Heav avy C Commercial al Veh ehicles es vs s Passe ssenger Ca Cars

➔ TOOL: part of b

broader er ec ecosystem em Trans nspor

  • rt sol
  • lution
  • n – mov
  • ving

ng p peop

  • ple

an and good

  • ods

Gen ener erating busi siness ss val alue an and profit it for

  • r ow
  • wne

ners - cust stomiza zable ➔ Long

  • ng l

life span; n; Second

nd l life; r

res esale e val alue ➔ High mileag age ➔ High h depend ndability; empha hasis on

  • n

de degrade ded d mode des. ➔ Highl hly m mod

  • dular:

➔ Low

  • w prod
  • duction
  • n vol
  • lumes; hi

high h variability / Emphasi sis s on D& D&D D cost sts

18

slide-18
SLIDE 18

Advan antag ages of

  • f autonom
  • nomou
  • us HCV

CVs

➔ Logist stics. s. Trucks currently limited in speed. ➔ Env nvironm

  • nment

ntal. Air resistance – convoying - Fuel savings ➔ Chauffer er r rel elated ed. Shortage of qualified drivers Truck dri river > r >33% 3% in in cost st ➔ Sim implif ific icatio ion ( (eventual) Stressful job and environment regulations to help drivers Design to help the driver: ergonomics, ➔ Ne New busi siness m ss models s possible if “C” drivers license is not

  • essential. Lower cost of entry for more

people. Source: Sveriges Åkeriföretag

33 % 33 %

19

slide-19
SLIDE 19

Sa Safety cons nsiderations ns specif ific ic t to HCVs

ALARP; ISO26262

  • Are

re curre rrently dri riven by pr prof

  • fessiona
  • nal drivers.
  • Coul
  • uld c

carry Haz azMat at

  • The

e size o e of H HCV, num number of

  • f pe

peopl

  • ple

trans nspor ported i inc ncreases the p possib ibilit ility a and s scale le

  • f d

dam amag age.

20

slide-20
SLIDE 20

Ou Outline

➔ Ab About the e Author(s (s) ) and nd the he projec ect ➔ Bac Background ➔ Complexity an and Legac acy impac act ➔ Advant ntages of

  • f a

autonom

  • nomou
  • us H

HCVs vs passen enger er veh ehicles es

➔ Cases: A ADI i int ntegration.

  • n.

➔ Conc

  • nclusions
  • ns, future wor
  • rk

and nd questions

  • ns

21

slide-21
SLIDE 21

Foc

  • cus on
  • n

per erspec ectives es o

  • f

➔ Busi siness A ss Asp spects ➔ Sa Safety ➔ Dependabil ilit ity ➔ Verif ific icatio ion ➔ Real alizat ation

23

slide-22
SLIDE 22

So Sources of

  • f ou
  • ur

challen enges es

➔Drastic increase in essential complexity ➔Socio technical implications that arise due to the potentially disruptive nature

  • f autonomy

➔The absence of a driver to deal with unexpected failures. ➔Saf afety av avai ailab ability trad adeoff

24

slide-23
SLIDE 23

Why cases at all?

➔ Man any skew ewed ed disc scussi ssions ➔ Exper ertise a e and differ eren ent cons

  • nsiderations
  • ns i

in n play. ➔ Legac acy i is a a moving t tar arget ➔ Prot

  • tot
  • type v

vs prod

  • duct

SO SOTA: mor

  • re or
  • r l

less prot

  • tot
  • types;

OEM EM IP IP ➔ Reluct ctance ce, c cos

  • st to
  • cha

hang nge legacy : ne needs m mot

  • tivation.
  • n.

➔ All cas ases ar are cap apab able of L5 au automat ation

25

slide-24
SLIDE 24

Sc Scope and delim imit itatio ions

➔ Issu ssues s common to all c case ses s e.g. col

  • llabor
  • ration
  • n with

h ot

  • the

her entit itie ies, l legal is issues, lia iabil ilit ity ➔ Foc

  • cus is on
  • n how

how the he ADI can n inte tegrate te with th th the platf tform ➔ Enabling reu euse e (wher ere e fea easible, e, r rea easonable, e, practic ical) is is a prio iorit ity.

26

slide-25
SLIDE 25

Mai ain contrib ibutio ions

To

  • the

he b best of

  • f ou
  • ur know

nowledge, this is paper is is uniq ique in in ➔ int ntegrating ng i int ntelligenc nce to

  • an

n autom

  • mot
  • tive p

platfor

  • rm

➔ Di Disc scussi ssion across ss su such a a broad ad sp spectrum of

  • f asp

spects w.r .r.t. autonom

  • nomy

Autonom

  • nomy mi

mindma map

27

slide-26
SLIDE 26

ADI d defin init itio ion

By Patrick Edwin Moran (Own work) [GFDL (http://www.gnu.org/copyleft/fdl.html) or CC BY 3.0 (http://creativecommons.org/licenses/by/3.0)], via Wikimedia Commons

➔OODA loop

  • op; O

Obs bserve, Orient nt, Decide de a and A nd Act. ➔Orien ent and D nd Decide de di directly mapped d to t

  • the

he ADI DI ➔Obs bserve a and A nd Act m mapped d to bot

  • both

h the he p platfor

  • rm and

nd the he A ADI, sens nsor

  • rs ne

need t d to

  • be r

e reu eused ed 28

slide-27
SLIDE 27

Con

  • ndition
  • ns f

for

  • r

reuse se

➔ Safety a ana nalysis depe pend nds on

  • n

conf

  • nfigur

uration,

  • n, coul
  • uld cha

hang nge pe per case a and nd c cont

  • ntext.

➔ It cannot nnot be avoi

  • ided, how

however the e need eeded ed analysis could be e min inim imiz ized. ➔ Legacy c com

  • mpone

ponent nts can n be reus used onl

  • nly if

Usage age still m meets de design de n decisions

  • ns bot

both h timing ng, and da nd data l limitations

  • ns.

29

slide-28
SLIDE 28

Assu ssumptions

➔ Com

  • mpone
  • nent

nts can b n be t turne ned

  • ff i

if need eeded ed ➔ Fai ail Saf afe vs Fai ail Operat ational al. ➔ Actu tuato tors limite ted to to th the plat atform ➔ New ew Sen ensors can b be a e added ed to the e ADI freel eely ➔ ADI c can an ac access al all inf nfor

  • rmation
  • n available to
  • the

he com

  • mpone
  • nent

nt it c cont

  • ntrol
  • ls

30

slide-29
SLIDE 29

Key Go Goals

➔Hig ighly ly S Safe, de dependa ndabl ble p platfor

  • rm.

➔Ea Ease of

  • f te

testi ting ➔Low v varia iabilit ility ➔Re Reuse of

  • f lega

gacy is a s a prio iorit ity

31

slide-30
SLIDE 30

Ca Case 1 1

Extrem eme; e; Ridiculous; Nec eces essary d del elimiter er

Source: By Humanrobo (Own work) [CC BY-SA 3.0 (http://creativecommons.org/licenses/by-sa/3.0)], via Wikimedia Commons

32

No Reuse Full Reuse

Case 1 Case 2 Case 3 Case 4 Case 5

Preserve legacy Optimize functionality

slide-31
SLIDE 31

Ca Case 2 2

Prot

  • tot
  • type

pes; Easi siest st

33

No Reuse Full Reuse

Case 1 Case 2 Case 3 Case 4 Case 5

Preserve legacy Optimize functionality

slide-32
SLIDE 32

Ca Case 3

Prot

  • tot
  • type

pes; Refine ned c cont

  • ntrol
  • l

34

No Reuse Full Reuse

Case 1 Case 2 Case 3 Case 4 Case 5

Preserve legacy Optimize functionality

slide-33
SLIDE 33

Ca Case 4 4

Traditiona

  • nal method

hods, conc

  • ncrete s

sol

  • lution
  • n

35

No Reuse Full Reuse

Case 1 Case 2 Case 3 Case 4 Case 5

Preserve legacy Optimize functionality

slide-34
SLIDE 34

Ca Case 5 5

The e other er ex extrem eme; e; Delim imiter

Int ntent ntiona nally le left bla lank!

36

No Reuse Full Reuse

Case 1 Case 2 Case 3 Case 4 Case 5

Preserve legacy Optimize functionality

slide-35
SLIDE 35

Resu sults: s: Pres eser erve L e Leg egacy ap approac aches

1 2 3 4 5

Higher Platform Reuse Lower accidental Complexity(on reuse) Lower Variability(acro ss platform) Lower Development Cost Upfront Lower Development Cost over time Higher Reliability/Avail ability Reduced need for Diagnostics to ensure f t Higher Security Ease of Verifcation of Modified Pf functionality Ease of Verification of ADI functionality Lower Information flow needed and…

Case 1 Case 2 Case 3

Platform Re Reuse Lim imit ited a accid idental Complex exity o

  • n

n reu euse Lower V Varia iabil ilit ity Lower er D Dev evel elopment ent Cos

  • st U

Upfron

  • nt

Lower er D Dev evel elopment ent Cos

  • st ov
  • ver t

time Higher her Relia iabil ility/ Avail ilabil ilit y Min inim imal D Dia iagnostic ics Higher her S Sec ecurity Ea Ease o

  • f Verif

ifcatio ion of

  • f

Modif ifie ied Pf Pf functio ionalit ity Ease of V Verif ific icatio ion

  • f A

ADI f functio ionalit ity lower er I Inf nformation n flow need needed ed and nd infra rastru ructure re

37

slide-36
SLIDE 36

Results: Optim imiz ize for functio ionalit ity approaches

1 2 3 4 5 Higher Platform Reuse Lower accidental Complexity(on reuse) Lower Variability(across platform) Lower Development Cost Upfront Lower Development Cost

  • ver time

Higher Reliability/Availabi lity Reduced need for Diagnostics to ensure safety Higher Security Ease of Verifcation of Modified Pf… Ease of Verification of ADI functionality Lower Information flow needed and infrastructure Case 4 Case 5 Platform Re Reuse Lim imit ited a accid idental Complex exity o

  • n

n reu euse Lower V Varia iabil ilit ity Lower er D Dev evel elopment ent Cos

  • st U

Upfron

  • nt

Lower er D Dev evel elopment ent Cos

  • st ov
  • ver t

time Higher her Relia iabil ility/ Avail ilabil ilit y Min inim imal D Dia iagnostic ics Higher her S Sec ecurity Ea Ease o

  • f Verif

ifcatio ion

  • fModif

ifie ied Pf Pf functio ionalit ity Ease of V Verif ific icatio ion

  • f A

ADI f functio ionalit ity Lo Lower Inf nformation n flow need needed ed and nd infra rastru ructure re

38

slide-37
SLIDE 37

Results: Case 3 vs Case 4

1 2 3 4 5

Higher Platform Reuse Lower accidental Complexity(on reuse) Lower Variability(across platform) Lower Development Cost Upfront Lower Development Cost over time Higher Reliability/Availabilit y Reduced need for Diagnostics to ensure safety Higher Security Ease of Verifcation

  • f Modified Pf

functionality Ease of Verification

  • f ADI functionality

Lower Information flow needed and infrastructure

Case 3 Case 4

39

slide-38
SLIDE 38

Fin indin ings an and Conc nclusions ns

➔ High variant p platforms & & ISO 26262 26262 = Cha halleng nge ➔ Com

  • mpone

ponent nt reus use is not not trivial w whe hen n sa safety i is c s consi sidered ➔ ADI => mor

  • re featur

ure int nteraction.

  • n.

Caref eful m managem emen ent req equired ed. ➔ ADI and nd pl platfor

  • rm ne

need to

  • evol
  • lve

toget ether er

Or risk Fail s safe be beha havior

  • r a

and nd low

  • w

de dependa ndabi bility

➔ Need for

  • r com
  • mpa

partment ntalization

  • n and

nd pa partition

  • n the

he ADI in n all cases. For

  • r

saf afety an and v verificat ation. ➔ Redund undanc ncy is key for

  • r hi

highe her de dependa dabi bility

40

slide-39
SLIDE 39

Fut utur ure wor

  • rk an

and proje jects started

For

  • rmalization/
  • n/ r

remov

  • ving

ng ambigui uity

  • Ont

ntol

  • log
  • gy of
  • f terms in

n our

  • ur spe

pecific cont

  • ntext
  • Use

e of an ea earlier er architec ecture e rec ecover ery pr proj

  • ject t

to r

  • refine

ne d defini nitions

  • ns of
  • f the

he layers in t n the he pl platfor

  • rm, the

he c cases, rul ules for

  • r reus

use Grand nd Coope

  • operative D

Driving ng C Cha halleng nge case se st study, STPA bas ased ap approac ach; C Cas ase 2. Systems thi hink nking ng ICES ind ndus ustrial ne networ

  • rk ASAP wor
  • rkgroup
  • up

wor

  • rkshop

hop is being ng pl planne nned.

41

slide-40
SLIDE 40

Ta Take aw away ays

Other than REA EAD THE E PAPER ER FOR MORE DETAILS!

  • Aut

utonom

  • nomy is e

essent ntial f for

  • r HCVs.
  • Prot
  • tot
  • type

pe vs pr prod

  • duc

uct

  • Sa

Safe state trans nsition

  • n ha

has to

  • be

guaranteed, id ideally lly w wit ith formal l verif ific icatio ion

  • Deg

egraded ed modes es are e critical in the e absenc nce of

  • f the

he hum human. n.

  • Deep

p int ntegration

  • n of
  • f ADI with

h pl platfor

  • rm

is n need eeded ed.

  • High variant p

platforms & & ISO 26262 26262 = Cha halleng nge

Cont

  • ntact: Naveen

eenm@ kth.se KT KTH- MECHATRONICS -ARCHER

ER

42