AU­TON­O­MOUS CARS NEED TO OVER­COME MA­JOR HUR­DLES BE­FORE THEY BE­COME MAIN­STREAM

In an ex­clu­sive in­ter­ac­tion with Dataquest, Ram Ramaseshan, Senior VP and Head, Au­to­mo­tive and Industrials BUs, Sasken Tech­nolo­gies, talks about the chal­lenges faced by au­ton­o­mous cars in In­dia and how can In­dia speed up to­wards Au­ton­o­mous driv­ing. Ex­cerp

Dataquest - - FRONT PAGE -

What is so spe­cial about your of­fer­ings and how does it help your clients? Sasken’s of­fer­ings help our cus­tomers ac­cel­er­ate time-to-mar­ket, en­abling adop­tion of tech­nolo­gies de­manded by con­sumers, and achieve cost ad­van­tage so that th­ese so­lu­tions can be brought to emerg­ing mar­kets at com­pet­i­tive price points. To­wards this, Sasken is fo­cus­ing on the fol­low­ing of­fers:

Cock­pit do­main con­troller us­ing Vir­tu­al­iza­tion:

In­creas­ing cus­tomer de­mand to bring con­sumer elec­tron­ics fea­tures to pas­sen­ger cars is lead­ing to pro­lif­er­a­tion of ECUs and ex­po­nen­tial in­crease in com­plex­ity of the car’s cock­pit de­sign. Num­ber of ECUs in high-end cars has in­creased from close to 40 dur­ing late 2000 to 100+ in re­cent times. This has de­cel­er­ated the rate of in­no­va­tion and fea­ture in­tro­duc­tion, in­creased R&D costs and chances of fail­ures for both Auto OEMs and Ti­er1s.

ECU con­sol­i­da­tion is an emerg­ing trend that al­lows mul­ti­ple ECUs to be con­sol­i­dated into a sin­gle SoC lead­ing to cost ad­van­tage of high end cock­pit Do­main sys­tem.

ADAS val­i­da­tion

Sasken is build­ing a sim­u­la­tion frame­work for val­i­da­tion of Ad­vance Driver as­sist sys­tems. Cur­rently test­ing of vi­sion-based ADAS sys­tems re­quires huge amount of an­no­tated video data for com­pre­hen­sive sce­nario cov­er­age. The gen­er­a­tion of this video con­tent for var­i­ous ge­ogra­phies, ter­rain and weather con­di­tions and traf­fic sce­nar­ios and the sub­se­quent man­ual an­no­ta­tion of the same for al­go­rithm val­i­da­tion is a pain point in the in­dus­try to­day

Sasken’s so­lu­tion will en­able gen­er­a­tion of test video data based on a sce­nario de­scrip­tion along with sim­u­lated sen­sor sig­nals to test vi­sion-based and sen­sor fu­sion al­go­rithms against. The so­lu­tion will also com­pare the al­go­rithm out­put against the sce­nario de­scrip­tion for ac­cu­racy and la­tency of de­tec­tion. Sasken’s pa­ram­e­ter­ized so­lu­tion al­lows mod­el­ling of var­i­ous drive sce­nar­ios with com­bi­na­tion of var­i­ous drive con­di­tion pa­ram­e­ters (e.g., bright­ness, en­vi­ron­men­tal con­di­tions, pedes­tri­ans, ob­sta­cle, etc.) This helps cre­ate a rich test bed video that can be used for val­i­da­tion of the ADAS al­go­rithms.

What are the chal­lenges faced by au­ton­o­mous car in­dus­try?

Au­ton­o­mous cars have some ma­jor hur­dles to over­come be­fore they be­come part of the trans­porta­tion land­scape. There are fol­low­ing key chal­lenges faced by au­ton­o­mous car in­dus­try in to­day’s world: Cost: How to achieve economies of scale so that tech­nol­ogy moves to mid tier seg­ments where vol­umes are there

Stan­dard­iza­tion: Lack of in­ter­op­er­abil­ity for Radar, Lidar, Vi­sion data in car net­work

Reg­u­la­tory: Reg­u­la­tion for Au­ton­o­mous cars are evolv­ing. There is no clear stan­dards for test­ing, cer­ti­fi­ca­tion of au­ton­o­mous cars, safety reg­u­la­tions, def­i­ni­tion on “Who is li­able in case of a mishap…”

Val­i­da­tion: lengthy val­i­da­tion cy­cles. As per the re­port from Rand cor­po­ra­tion ti­tled, “How Many Miles of Driv­ing Would It Take to Demon­strate Au­ton­o­mous Ve­hi­cle Re­li­a­bil­ity?”, it would take 275 mil­lion miles/12.5 years of driv­ing to demon­strate with 95% con­fi­dence that the fail­ure rate is at most 1.09 fa­tal­i­ties/100 mil­lion miles. This is a long val­i­da­tion cy­cles and there is a need to adapt non-tra­di­tional ways to ac­cel­er­ate val­i­da­tion of au­ton­o­mous ve­hi­cles.

Ac­cord­ing to you, by when can we ex­pect au­ton­o­mous cars in In­dia? What se­cu­rity mea­sures do com­pa­nies take while build­ing th­ese cars? And how safe are th­ese cars for the pas­sen­gers?

Au­ton­o­mous cars have mul­ti­ple chal­lenges in In­dia as fol­lows:

In­fra­struc­ture: The Road in­fra­struc­ture is evolv­ing and still no way com­pa­ra­ble to what’s found in western world. Au­ton­o­mous ve­hi­cle re­quire high speed con­nec­tiv­ity which is re­stricted to ur­ban cen­tres since the Qual­ity of ser­vice in hin­ter­land is still patchy

Eco­nom­ics: The pri­mary driver for move­ment to­wards au­ton­o­mous ve­hi­cles in west is to­wards Ride hail­ing and ride shar­ing ap­pli­ca­tions. The cost of hir­ing a driver is ex­pen­sive in th­ese ge­ogra­phies. Con­trast this with In­dia It is Still in-ex­pen­sive to hire a driver. The cost of the ve­hi­cles in­cor­po­rat­ing th­ese tech­nolo­gies has to be com­pet­i­tive for adop­tion in high growth mar­kets like In­dia.

Net­work ef­fect: Au­ton­o­mous ve­hi­cle will be suc­cess­ful only if there a net­work of au­ton­o­mous ve­hi­cles and an evolved eco-sys­tem that co-op­er­a­tively drive.

Even ad­vanced na­tions are at least 5 years from de­ploy­ing au­ton­o­mous ve­hi­cles com­mer­cially in some shape or form though there are many tri­als hap­pen­ing. In In­dia the fo­cus on au­ton­o­mous cars will be less to­wards re­plac­ing labour but more to­wards as­sist­ing the driver in or­der to make roads safer.

In our view, we can ex­pect au­ton­o­mous ve­hi­cles can ap­pear in In­dia by 2025 that too at level 2 which cor­re­sponds to par­tial au­to­ma­tion – the driv­ing mode-spe­cific ex­e­cu­tion by one or more driver as­sis­tance sys­tems of both steer­ing and ac­cel­er­a­tion/ de­cel­er­a­tion us­ing in­for­ma­tion about the driv­ing en­vi­ron­ment and with the ex­pec­ta­tion that the hu­man driver per­form all re­main­ing as­pects of the dy­namic driv­ing task

Se­cu­rity is a ma­jor con­cern when it comes to au­ton­o­mous ve­hi­cles. The sur­face of at­tack of au­ton­o­mous ve­hi­cles has in­creased as the elec­tron­ics com­po­nents in the ve­hi­cle have in­creased. A Mul­ti­level se­cu­rity mea­sures along with con­sumer ed­u­ca­tion is re­quired to ad­dress the se­cu­rity is­sues.

Ac­cord­ing to the June 2015 early es­ti­mate of mo­tor ve­hi­cles fa­tal­i­ties from U.S. Na­tional High­way Traf­fic Safety Ad­min­is­tra­tion (NHTSA), an es­ti­mated 32,675 peo­ple died in car crashes in 2014. The Eno Cen­ter for Trans­porta­tion, a think tank, notes that “driver er­ror is be­lieved to be the main rea­son be­hind over 90 per­cent of all crashes” with drunk driv­ing, dis­tracted driv­ers, fail­ure to re­main in one lane and fall­ing to yield the right of way the main causes. Be­cause the ma­jor­ity of th­ese ac­ci­dents are caused by hu­man er­ror, self-driv­ing cars could po­ten­tially re­duce the rate of au­to­mo­bile-re­lated deaths—and save the U.S. over $400 bil­lion (2 per­cent of the U.S. GDP) in to­tal an­nual costs of ac­ci­dents.

So, Au­ton­o­mous ve­hi­cle (ve­hi­cles with var­i­ous de­grees of au­ton­omy) are rel­a­tively safer than man­ual driv­ing. How­ever, users need to get used to be­ing driven by a com­puter and that change in mind­set will take time.

How can In­dia speed up to­wards Au­ton­o­mous driv­ing?

Govern­ment’s in­cen­tive for safer ve­hi­cles should be in­tro­duced so that it will fo­cus on the au­ton­o­mous driv­ing. Pos­si­bly a cer­ti­fi­ca­tion mech­a­nism in lines of BS for emis­sion for safety would in­cen­tivise ve­hi­cle man­u­fac­tures to in­cor­po­rate some of th­ese tech­nolo­gies.

Set­ting up of ded­i­cated lanes/tracks where the ve­hi­cle can be tested – the smart city ini­tia­tive of the Govt. of In­dia would be a good pre­cur­sor to this and the ‘smart cities’ would be good can­di­dates for im­ple­ment­ing in­fra­struc­ture for au­ton­o­mous ve­hi­cle test­ing. Busi­ness mod­els such as ‘Can In­dia be test bed for Asian/African mar­ket?’ would in­cen­tivize more pri­vate-pub­lic part­ner­ships in this area.

—RAM RAMASESHAN Senior VP and Head, Au­to­mo­tive and Industrials BUs, Sasken Tech­nolo­gies

Newspapers in English

Newspapers from India

© PressReader. All rights reserved.