SPE­CIAL RE­PORT

Auto components India - - CONTENTS - Story by: Sricha­ran R

High tax rate stunts auto in­dus­try po­ten­tial, says SIAM Pres­i­dent MathWorks is fu­ture-ready with prod­ucts for new mo­bil­ity Grund­fos to launch new pumps for au­to­mo­tive in­dus­try

MathWorks, a lead­ing de­vel­oper of math­e­mat­i­cal com­put­ing soft­ware, has its pro­grammes ready for the fu­ture of mo­bil­ity in­clud­ing au­to­mated driv­ing and elec­tric ve­hi­cles. The US-based soft­ware com­pany re­cently con­ducted a sem­i­nar in Chennai, fo­cus­ing on sys­tem de­sign, de­vel­op­ment, and anal­y­sis us­ing MATLAB and Simulink. The ses­sion fo­cused on up­com­ing tech­nolo­gies in the in­dus­try like data an­a­lyt­ics, IoT, au­to­mated driv­ing, EVs etc.

“EVs and au­to­ma­tion are the trends we are see­ing. Ev­ery com­pany, OEM, and Tier - I sup­pli­ers are work­ing on that. Ev­ery au­to­mo­tive or­gan­i­sa­tion is now fo­cussing on th­ese 2 trends. We are see­ing this as a huge in­ter­est com­ing and we are try­ing to en­gage with them,” Vi­jay­alayan R, Man­ager of Con­trol De­sign Ap­pli­ca­tion En­gi­neer­ing, MathWorks, told Au­toCom­po­nentsIn­dia.

He fur­ther said, “We will have the same prod­ucts for the EVs as well. Let it be an elec­tric ve­hi­cle or a con­ven­tional pow­er­train, the work­flow re­mains the same. In Pow­er­train Block­set, we have added some ref­er­ence ap­pli­ca­tion for EVs and hy­brids. This brings the pos­si­bil­ity of hav­ing an ar­chi­tec­tural model there. The cus­tomer can use this as a ref­er­ence to put their data and make a par­tic­u­lar model, use it for the en­gi­neer­ing de­ci­sions. It also helps them to sim­u­late at a spe­cific drive and ar­rive at a con­clu­sion.”

A Pow­er­train Block­set, re­leased in 2016 pro­vides fully as­sem­bled ref­er­ence ap­pli­ca­tion mod­els of au­to­mo­tive pow­er­trains, in­clud­ing gaso­line, diesel, hy­brid, and elec­tric

sys­tems. It in­cludes a com­po­nent li­brary for sim­u­lat­ing en­gine sub­sys­tems, trans­mis­sion as­sem­blies, trac­tion mo­tors, bat­tery packs, and con­troller mod­els. The prod­uct also pro­vides a stan­dard model ar­chi­tec­ture that can be reused through­out the de­vel­op­ment process. Sim­i­lar to this, many such prod­ucts were brought in by the soft­ware firm to help and ease en­gi­neers in the au­to­mo­tive in­dus­try.

Speak­ing on the new prod­ucts they have brought into the mar­ket for the cur­rent trends, Vi­jay­alayan said, “It is not just about new prod­ucts. What we have ob­served from the In­dian and the global au­to­mo­tive in­dus­try is that a ma­jor trans­for­ma­tion like the au­to­mated driv­ing, elec­tri­fi­ca­tion, fleet data an­a­lyt­ics etc is hap­pen­ing. Be­cause of th­ese, we see mul­ti­ple in­dus­tries com­ing to­gether. What used to be only an au­to­mo­tive sec­tor is now a con­ver­gence of in­dus­try. For ex­am­ple, be­cause of au­to­mated driv­ing, we see semi­con­duc­tor com­pa­nies get­ting into that space. For us, the con­nect­ing point is that we

pro­vide a plat­form in which mul­ti­ple in­dus­tries work to­gether. What we try is to pro­vide a plat­form for sys­tem sim­u­la­tion, per­form­ing data an­a­lyt­ics and an en­vi­ron­ment in which they can test, and de­velop the al­go­rithm which is mainly for the au­to­mated driv­ing. We call this ver­i­fi­ca­tion and val­i­da­tion. An au­tomaker will not get the time to go to the field and do the test­ing each and ev­ery time. That is where they rely on meth­ods like sim­u­la­tion to cre­ate a vir­tual test ground to ver­ify the al­go­rithm.”

MathWorks has re­leased up­dated prod­ucts and also added ca­pa­bil­i­ties to the ex­ist­ing prod­ucts. “With au­to­mated driv­ing com­ing into play, there is a need to fo­cus on ve­hi­cle dy­nam­ics. For this, we launched Ve­hi­cle Dy­nam­ics Block­set in 2018, that fo­cusses on mod­el­ing the ve­hi­cle dy­nam­ics both lon­gi­tu­di­nal and lat­eral. This will help the en­gi­neers to test the chas­sis con­troller or even the au­to­mated driv­ing al­go­rithm. With this, we also pro­vide an in­te­gra­tion with a gam­ing en­gi­neer called Un­real, where we can cre­ate sce­nar­ios with the road, etc. So one can mount the sen­sor on the ve­hi­cle and test it in Matlab sim­u­la­tion and at the same time wit­ness it in a 3D world also.” On the other trends of the au­to­mated driv­ing, Vi­jay­alayan said that there are 3 chal­lenges that the cus­tomers face. First, the au­to­mated driv­ing de­pends on sen­sors. Sec­ond, deal­ing with a lot of videos and third, the sce­nario cre­ation. With the un­der­stand­ing of this, they came up with an­other prod­uct called Au­to­mated Driv­ing Sys­tem Tool­box that pro­vides al­go­rithms and tools for de­sign­ing and test­ing ADAS and au­ton­o­mous driv­ing sys­tems. One can au­to­mate ground-truth la­bel­ing, gen­er­ate syn­thetic sen­sor data for driv­ing sce­nar­ios, per­form mul­ti­sen­sor fu­sion, and de­sign and sim­u­late vi­sion sys­tems with this. “The chal­lenge for the en­gi­neers is ba­si­cally how do they vi­su­alise the in­for­ma­tion com­ing from the sen­sors and ex­tract­ing the sights out of this. Next, they will be deal­ing with a lot of videos that are to be part of the in­tel­li­gence sys­tem. For that, they have to make a per­cep­tion al­go­rithm. So, to make that you need to get the real time sce­nar­ios from the videos while do­ing videos one need to la­bel the ob­jects that are present in the video and that is what is made into the al­go­rithm. This is a very time-con­sum­ing work. The sce­nario cre­ation in au­to­mated driv­ing is all about how the al­go­rithm can work for the edge cases also. Our prod­ucts pro­vide a ref­er­ence plat­form for per­form­ing sen­sor fu­sion, syn­thetic sce­nario gen­er­a­tions and also vi­su­al­is­ing the sen­sor data. It will also be the ref­er­ence plat­form and make sure that the en­gi­neers do not need to go to the ba­sics to build the al­go­rithm, as they have a quick ref­er­ence model avail­able. In this way, we are try­ing to re­duce the de­vel­op­ment and test­ing time,” he said.

Adding more, he said that the data an­a­lyt­ics is where they trend now. There are a lot of data avail­able in the mar­ket with the OEMs and the

Tier-I sup­pli­ers. The ques­tions are how to ex­tract data from the in­sight ma­te­ri­als. As they are huge it will be tough han­dling them. That is where they are in need of AI.

“There is a short­age of data sci­en­tists and with Mathworks, we are go­ing to pro­vide the ca­pa­bil­i­ties that en­able en­gi­neers to turn as a data sci­en­tist. So, if they wanted to learn ma­chine learn­ing or deep learn­ing we pro­vide them apps that are user-friendly. In 2018, we came up with a new pre­dic­tive main­tain tool­box, where we use the sim­u­la­tion mod­els to sim­u­late the fail­ures and get a data out of it. So one can use the mod­els to gen­er­ate re­sults by in­ject­ing fail­ures,” he stated.

Nowa­days, sim­u­la­tion plays a ma­jor role in the R&D sec­tor. They use sim­u­la­tion to make sure that their ideas and con­cep­tu­al­i­sa­tions work bet­ter. The spokesper­son states that, “Sys­tem level sim­u­la­tion oc­cu­pies a ma­jor role in con­cept prov­ing and just by look­ing at a text one can­not de­cide whether it is fea­si­ble or not. It helps the peo­ple of R&D to

con­cep­tu­alise their ideas in forms of ex­e­cutable mod­els and they can know whether this works or not. With this, they can de­velop and im­prove a lot more.”

Speak­ing about the ve­hi­cle life­cy­cle th­ese days, he said, “What we see is, a time taken for build­ing a ve­hi­cle has dras­ti­cally come down over the pe­riod of time. The main chal­lenge is the devel­op­men­tal time and the amount of soft­ware that has gone into a ve­hi­cle have in­creased. What used to be a 5% of soft­ware in the ve­hi­cle has now be­come around 20-25% and it keeps grow­ing. The ques­tion is also for the man­u­fac­tur­ers and Tier-I sup­pli­ers, on how do they han­dle the par­tic­u­lar chal­lenge of meet­ing the shorter de­vel­op­ment time, and also keep adding a new fea­ture to your ve­hi­cle. This is where the model-based de­sign is help­ing. So, when they make a model for one ve­hi­cle tool, most of the func­tions can be re­us­able for the other ve­hi­cle de­vel­op­ment. Reusabil­ity is im­por­tant. And now, it is no more a tra­di­tional hand cod­ing.

Model-based de­sign­ing is where I can make a model of the ve­hi­cle and it is also where mul­ti­ple things work to­gether and in­te­grate. That is where Simulink plat­form is re­quired. The same model can also be reused for the code gen­er­a­tion. In­stead of you writ­ing the codes it can au­to­mat­i­cally gen­er­ate the code from the model and de­ploy it. This saves an im­mense amount of time.”

On the 100% of ve­hi­cle sim­u­la­tion by en­gi­neers, VI­jay­alayan said, that it can­not be fully vir­tual. He feels that the time of the process can be re­duced and that will not elim­i­nate the phys­i­cal pro­to­type. To val­i­date a par­tic­u­lar model one need real data, he said. Speak­ing about the fu­ture, he stated that the soft­ware firm wanted to give a plat­form to the au­to­mo­tive or­gan­i­sa­tion in all trends and work closely with cus­tomers. “By work­ing closely we get feed­back, and that will help us in en­abling the en­gi­neers across the in­dus­try to lever­age the use of our tools and make them work faster, smarter,” he signed off.

Vi­jay­alayan R, Man­ager of Con­trol De­sign Ap­pli­ca­tion En­gi­neer­ing, MathWorks

Newspapers in English

Newspapers from India

© PressReader. All rights reserved.