Automatic Vehicle
`
`Guidance:
`
`The Experience of the ARGO
`~ u t < rgmo is Vehicle ' .
`Autonomous Vehicle
`
`'
`
`Alberto Broggi
`
`Massimo Bertuzzi
`
`:‘XIt-bsdndrd Faaciuli
`
`Gianni Cflnle
`
`
`
`
`
`
`
`.
`_,.‘
`ir'.‘
`
`I
`i
`..
`.
`‘Hu-I" .
`,
`-‘r-"'*"'
`.
`‘1'.“1-
`
`'-
`
`.
`
`_
`..'-'
`
`'
`
`-
`
`1...
`n-
`
`H "
`
`_
`
`I
`
`.1
`
`I
`I '
`.'.'
`I:
`'-
`--
`'_..
`..
`'
`-"':.1."'-'
`," EU'” '
`'H '
`.ar"..
`~
`..
`. Tunic
`" w
`.. "'11:?” wpr-
`z..:.'=:"r~w..~ "fl
`..
`._-
`gh$1x~fg1 f:'-'::'-I.f-Ff m¥t~5_
`4:
`.
`.
`ul_
`‘h-
`"#1-
`I
`I
`'
`.m-tfi.
`.
`“.afi
`a.
`ti.” .rf'k“ "."1-
`."_
`.‘r'I‘I'J
`;£‘-l
`“Q":
`"Tamil-4.1:.
`fin.
`"'
`-’-E -
`
`World Sclentlflc
`
`VALEO EX. 1009
`
`VALEO EX. 1009
`
`

`

`“amen-u:
`
`_' murmur-Imus
`,I‘Ffl- H- II. Immmmmmm
`[an nmummmmm
`
`
`
`
`“MW”
`AMI-Ind unnummumm.
`
`mmmmm
`tumult-mum
`
`Walmuwummmmm
`
`Mummmmrqmmdm
`mun—nuuumm-«mmmmm
`
`mmdmhhmH—n-mhwfl-W
`“Mk“!IIMMD-MHANIHLMhflmfl-h
`“autumn-mm.
`
`Ill“ HIM-m
`
`MMkMMIfl-hfiu
`
`Primal in Siam by Una-Print
`
`
`
`{undumnbymnfludbm
`they mdmd broad-flowing Plush,
`and the utmost hourna aim ma."
`
`Apallunlun, THE Aanumuncn. Book [I
`
`

`

`Intelligent Vehicles and Machine Wrists
`
`Recoil-mam of Intelh‘geal Transportation-I System
`
`13
`
`functionallties must he froaen and integrated in a fully optimized and engi-
`neered embedded system before marketing. It is in this stage of the project
`that the development. of ad-hoc custom hardware still plays a fundamental
`role and its costs are justified through a large scale market.
`
`2.2 Requirement: of Intelligent Transportation System
`
`hny on-board system for ITS applications that is to be sold on the market
`needs to meet some important requirements which are summarised in the
`
`The final system that will be installed on a commercial vehicle
`needs to be robust with respect to a number of different aspects.
`First of all. any automatic system —even a very simple one such
`as an intelligent cruise control. which controls only the vehicle’s
`speed~ must be smart enough to adapt to different conditions of
`environment [flat or hilly paths]. road (highways, extra-urban. or
`urban scenarios], traffic {dlfiesent volumes of vehicles and obsta-
`cles}, illumination (day. night. sunset. sunrise}. and weather {sunI
`fog. rain, or even snow]. and to tolerate nnexpemed changes of their
`parameters. The system must minimise —er better stlll reduce to
`aero— tlte mnnber of incorrect results signals since they may alters
`the user‘s confidence over the system, and induce the driver to telte
`over the commamhof the automatic system aodfor switch it ofl'.
`Meteor-Her, not only does the algmithnm’ robtntnem have to be
`granted. but also the hardware system —semors induded— requires
`the rmietance to enchanieal and tbermic stems, such as vibrations
`
`Production and Operative Costs
`It has been estimated that for marketing reasons as ITS system
`should cost no more than "1% of the vehicle price. These cost is-
`sues press-t no problems for truths, boom, or higb‘cost vehicles in
`general, tor vrhich the installation of an expensive ITS system can
`be regarded as an investnrerrton therebiele's security and safety.
`conversely. for the car market the cost issues heoorns fuadmuen-
`tally important. requiritq a specific engineering phase 1which may
`also need the redesign of the complete hardware platlorm. How-
`ever, the developmenr costs due to the system ee-migiueering will
`be compensated by the large car market.
`Furthermore, notorlly do produflloaeostsofl'I'Sappsrata need to
`be carefully considered and reduced as much as possible, but the
`operative costs need to be heptlow aswell. Power consumption 'at
`infect mtherley parameterthatnmstbeomeidered forthecar
`marl-net, since the vehicle's performance should not be altered by
`the use of ITS apparata.
`
`Sine and Design
`The roq'll'rernentsof the our market are very specie: notmly must
`the vehicle's performance not be altered and ITS production costs
`kept low, but also ear stylingshoold not sitier from the istallatlon
`of new hardware and —itt particular- new settlers for ITS- applica-
`tione. Therefnrcthese systonsmedtohecornpact in siseand the
`eensrnsneedtohe imalled in spoeition thatdoesaot causedis-
`tnrbance. Although for many swam their positioning does not
`present aesthetiml problems (for example, radars are typically po-
`flI-ioaed on the vehicle's iota: bumpers}. the httallulort of cameras
`may be considered as an additioam problem: some epgnoarhes re-
`miss the cantoras to he imtsled inside the driving cabin behind
`
`

`

` “
`
`Intel-figs!“ Vehicles ell-d Hachitts His-lott-
`
`Muei'l'ris Vision.
`
`15
`
`are also expected to become the front-end lfor ITS systems too.
`Hrrthermore. the use of visual signah {leds.
`lights. control pan-
`els. or even on-board monitors]. mediator-a] feedbaclts {steering
`wheel. seat. or pedal vibrations]. and vocal mmsages mad as driv-
`ing assistance warnings must all be carefully evahtated. The system
`especially a driving assistance one— most not flood the driver with
`a large quantity of iniormation since this mold reduce the driver's
`attention and cause nothing else but problems.
`
`1.3 Sensing the Environment
`
`The main requirement of ITS systems is a way of sewing the surround-
`ing envirmnrent. Five different ways to perceive inibrmation about the
`environment commonly used in indoor robotics are tactile. acoustic. laser,
`radar. and vision sensors.
`
`Tactile sensors are generallyr used in indoor robotics. where the
`speech: of robots are sufficiently lov-+ so as not to cause serious dame
`ages 1when hitting obstacles. Obviously in automotive applications
`tactile scissors. such as bumpers. are of no practical use.
`Moutic sensors
`
`Acoustic sensors are interesting mainly because of their low cost.
`but. due to their very reduced detection range. are seldom'used
`in automotive applications. These kind of sensors are classified
`as active sensors. became their use involves the measurement of
`alterations of signals emitted by the sensors thesaselves.
`Laser-based. sensors
`
`Based on the Doppler sfl'ect. laser-based radars detect the d'mtsnce
`
`based radars. Unisrtunately. both of them also share the same
`problemsc-E lowspat'ral resolution and slovr scanning weeds. Again.
`radar-hosed seams are classified as eel-ire sensors.
`Vision-laud sensor:
`
`1|Ii'ision-iteoed sensors are sessile sensors and their use has several
`advantagm over radar. laser. acoutic. and tactile cursors: for ex-
`ample the possibilityr to acquire data in a non-invasive tray. time
`not altering the environment.
`in this casethe scanningoftlta imageisperfiomtediast enough
`for 1T5 applications. Moreover. they can he mad [or some specific
`appflcations for which visual hEorreation plays a basic rule (such
`as:
`lane maritime heahatim. trsfic signs recognitim. sleuth
`identificathn} without any modifications to road infrestmdures.
`Uniortnnatel] vision sensors are tam robust than mflimetenwave
`radars in foggy. night. or sheet sun-shine ooniitions.
`
`2.4 Machine Vision
`
`The main pmblems due tntbeuse oi'activeseusorsubeeider the pollution of
`the environment— include interwehicle interference amongst the same type
`of sensors and the wide variation in refioc-Ilcei ratios coined by many difl'er-
`ent reasons. such as obstacles‘ shape or material. Moreover. the maximum
`signal lerel must comply with some safety rules and must he lovrer than a
`safety threshold.
`0n the other hand. the use of active acnsors. which involves the mea-
`surement of the alteration of signals emitted by the sensors themselves.
`possesses some specific peculiarities
`
`v active sensors can measure quantities in a more direct vray than
`vision.
`its an example. a Doppler raihr can directly measure the
`
`

`

`
`
`Mulligan! Vehicles arid Hachs'ne Vision
`
`Hardin's Vision
`
`1?
`
`piernentary tastes for the recognition of objects. detection of free-space, or
`in checking for certain specific object characteristics. Unfortunately, when
`several robots are moving within the canes environment, active sensors may
`interfere with each other. thus deer-Mag their reliability and usefulness.
`This problem becomes even greater in outdoor unstructured environments,
`in which a large number of vehicles could be moving shoultsneously. as
`rt'or example- in the case of autonomous vehiclm travelling on intelgent
`
`Hence. with a massive and widespread use of autonomous sensing agents.
`the use of passive sensors, such as cameras. obtains important advantages
`over active ones. These are the cases in which vision beer-rues ct' paramout
`flbvionely machine vision can fail —and thereibre cannot be
`used- when humans cannot see [c.g. in foggy conditions or during the night
`with no specific illumination]. In fact. while the use of other sensing agents
`would have the advantage of extending sensing capabilities besides human
`possibilities, the use of vision allows the building of s system able to act
`as the human driver: for enample. an active safety system that helps the
`driver when he fails {e.g. in the lack of concentration].
`
`The recent developments in computational hardware. such as a higher
`degree of integration and reduction of the poser supply voltage. allow the
`seem to machines that can deliver a high level of computational power.
`with fast networking facilities. at an sfiordsble price. Since the early stegm
`of v'asion [low-level image oncoming} are computationally demanding. the
`availability of lots-cost enginm helps to solve some basic bottlenecks Cur-
`rent terhnology allows the possibility of thfD-lihs proceming para-tights
`even in general-propose processors. such as the new generation of procre-
`sors that include multimedia extensicns [Visual hetrnction Set. V15. in
`UltreSPAftC, hfldli in Intel's Pentium, nix-e in Hewlett-Packard's PA-
`RISCS. MDMK in Silicon Graphics‘ MIPS, M‘r'l in Digitalis Alpha. 311le
`
`tiona can be found in GRIDS- tremors. anti as the possihiity of dealing with
`pixels independently as in tradtional memories. Another Itey advantage of
`ENDS-based sensors is that their integration on the process-grin]: seats
`
`to he straightforward.
`As a result. not only did this advanced technolog promote improved
`hard-rare devices. but also triggered oil" renewed interest in the techniqses
`for the. processing of iconic htfnrmstinl, germrslly sdtiessetl by the field
`of Artificial Intelligence. who deals with image interpretation and —more
`generaly— will perception, when the fission of data. coming from other
`more ls also lntegratei.
`Nonetheless. when designing a vision system for automotive apples-
`tions. some important charmerntirs must be carefully considered.
`
`- HS systems require taster moccsslag than other applications. since
`the vehicle speed is proportional to the processing rate. The main
`problem that has to be laced when real-time imaging is concerned
`
`andwhichlsintrinsictotheproccsaingol'll‘oagcs.lsthelarge
`amount of data —and therefore computation— involved. its a result,
`ape-rifle computer architecture! and presents-log techniques must. he
`do'rised in order to achieve real-time periormanoe. Nevertheless,
`since the success of ITS spperata is tightlyr minted to their cost.
`the computingenflnflelnnothehssedonexpemiveprocessors.
`‘I‘lwoefore. either olf-the-shelf components or sd-hoc dedicated low-
`cost solutions must he considered.
`
`a. Eithermore. in theatrtomotlvefield no assumptions can bemade
`onkeyparametersJormmpdetheillumilnthnorumtrsetclthe
`scene. which are directly measured by the vision seamr. Hence.
`monumentpcooesslngsnust herobustrmcsrghtotolu'stethese
`changw and both adopt to fies-em environmental conditions. such
`as sun [high hrightoem and cmtrnat due to shadows), rain [ex-
`
`

`

`
`
`BO
`
`Alpert-[hm fur “my! Penn-nut}
`
`ill-alum: fltlfl'llnl.‘ a He‘ll-Heed AM
`
`HI
`
`In
`The itigurithm luu nim- h-nrn t'lll‘l'lt'llfll tn hmttlle intest- means.
`this cue. a nutehle speed-up It Mllmnl ha. :tylmnitelly "timing the area
`til int-Hut INT. 1593. neulely the “mi-m where the Chen-cit! is summed.
`to be in the lull-wing frame ul' the qur-m.
`It
`in reduced to I. single
`image ‘l‘lfltltfl there thu allurin- have been pee-rim!!! detected;
`than!
`windmills; twinnqun E61151] dilue- I. he: Whfl‘ of the mating ehflaltle.
`The region of lute-Int in detenmnnl by telling the mailer bounding bu:
`including all the "lament-h lnfirnuun: in vehicle mean and the nearing
`1IIII'heIel angle cut be med to update thr- I'lltiltrl all utmtm m a. mute
`mum {Him-IL I'ltI'I-wl'f. mantra-lit ul when nu abated-en ere Emmi
`
`the unm- imfie It utnlylnl. In uni-:1 tn Ami-cl the rel: of missing new
`obtarha.
`
`4.2.2 Perfmnee andpeie
`
`The whole plenu- hle heen tilted an antigen mprmntiug set-Hui entra-
`urbnn duuhlthlmemede. with taint-1| {III-ll markings. Althtaugh the specific
`
`tam-em doe: not n-ly en the pmmrr ui maul mutiny. it has been tuned
`an Etrut'ttlred rmulr- since it hurl tn he integrand with the module aimed at
`
`the Identificatim ul lilflt' huutllhu'lel. mil thereiure hneed m: and ending
`detection. Unimtumltely. rnul makings tenement I. new feature that
`alt-we duwn tht'. mutium-lwei plume in lent, the number ”Manned mmem
`is increased M well an the lmtttllt‘r or rmulml "Melange.
`titan-erehut-lean}:1
`generally the rectum!“ fit‘ltl'lh‘l-Lt'tl
`luv uuctemd'ei lItMtfltmg tlfl nut comply
`with the size requirement mentimlml in Mllflll 4.2.1.2. Fur thin rem,
`the final mum ul mild rectunglm ltl net Hfll‘tfltxl by the prmenee at" reed
`
`markings. although mmlhly aluminum].
`Fimtre HT tare-lent: a ll." H! the result: regarding the Humming ef
`image: that Ihw a vehicle In: the haul:
`the mmlutinn Hi the hming
`grey—level inn-um: l! 256 u 256. B Intfpixel.
`The algulithm has been tested in Llilh-Ieut nmlitimu. without lit-uncles,
`on straight and tuned “It'll. in dim-rent hgllt mnditium. and an junctions-
`'l'he me critical behavior corrupted: m the penance nl eludm on the
`path.
`inducing. theeppeumulemtfldenhle numberofetlgce'm the
`
`Infill-eliminate the dew- mil Indium-level aim-reeling: the
`firmdwmdepmimthelpwtfirm'hihthe blah-trend]?
`mum-ed by them: dun-neat It'll-um {um}. Tnhletlmnfimu
`
`
`
`Flu, u? Sauna-plum the-union aim-Maw Ilene.
`
`Ihutunnflelfltmdcmedmheithmvehkhthelmfithmmluter
`lhenlnthemmnditmwlthmhieletmthepnh. 'I‘hh'udmtuthe
`mhudmuedhgrthe mmflu'tflfleflmphue. Dntheetherhuui.
`mflkuhkhfidmflthelmddetflltedguithemmherd
`
`