Led by Sor­bonne Uni­ver­si­ty, in an alliance with UTC, the PostGenAI@Paris pro­gramme is one of the 9 win­ners of the nation­al ‘IA-Clus­ter’ call for expres­sions of inter­est (CEI). Endowed with 35 Meu­ros over 5 years, this clus­ter will con­tribute to France’s strat­e­gy in arti­fi­cial intel­li­gence (AI), by cre­at­ing an inter­na­tion­al cen­tre of excel­lence specif­i­cal­ly ded­i­cat­ed to post-gen­er­a­tive AI.

This fund­ing will be used to sup­port research projects, to devel­op train­ing pro­grammes and to unite the aca­d­e­m­ic com­mu­ni­ty. In addi­tion to this pub­lic fund­ing, indus­tri­al funds will be used to strength­en part­ner­ships with­in col­lab­o­ra­tive accel­er­a­tion projects (CAP).

As a mark of recog­ni­tion and a dis­tinc­tion for UTC and its skills and exper­tise, the uni­ver­si­ty is involved in four CAPs and is lead­ing two of them. One by Véronique Cher­faoui for UTCHeudi­asyc (UMR-CNRS 7253) and its asso­ci­at­ed joint lab­o­ra­to­ry SIVALAB with Renault Group, the CAP ‘Twin­ning’ aims to study shared dri­ving between a dri­ver and a vehi­cle capa­ble of auton­o­my, in par­tic­u­lar by study­ing the inter­ac­tions between the dri­ver and the vehi­cle to enable autonomous, coop­er­a­tive and safe trav­el on open roads.

The sec­ond CAP, ‘Indus­try’, is co-direct­ed by Alexan­dre Durupt for UTC-Rober­val and Yves Grand­valet for UTC-Heudi­asyc, and also involves the joint DIMEXP lab­o­ra­to­ry with UTC-Rober­val and DeltaCAD.

And among the aims of this PAC? ‘The emer­gence of soci­etal issues, in par­tic­u­lar the eco­log­i­cal ques­tion, which is encour­ag­ing indus­try to be more effi­cient and more envi­ron­men­tal­ly vir­tu­ous. Hence the impor­tance of devel­op­ing method­olo­gies for error detec­tion and pre­dic­tive main­te­nance of indus­tri­al sys­tems. At present, all the recent AI tech­nolo­gies, and in par­tic­u­lar the advent of Deep Learn­ing, have made it pos­si­ble to devel­op more effec­tive fault detec­tion tech­nolo­gies, with ‘false pos­i­tives’ or false alarms need­ing to be as low as pos­si­ble,’ explains Alexan­dre Durupt.

Final­ly, the last two CAPs are sup­port­ed by Sor­bonne Uni­ver­si­ty. The first, with which the Bio­me­chan­ics and Bio­engi­neer­ing research unit (UTC-BMBI, UMR-CNRS 7338) is asso­ci­at­ed with Anne- Vir­ginie Sal­sac, and steered by Isabelle Bloch of Sor­bonne Uni­ver­si­ty, involv­ing clin­i­cal part­ners as well as numer­ous indus­tri­al part­ners. ‘We have estab­lished a very strong part­ner­ship with ANSYS France, which spe­cialis­es in mul­ti­physics dig­i­tal sim­u­la­tion. Our objec­tive at UTC is to fur­ther devel­op dig­i­tal mod­els of intra-car­diac blood flow and valve and ves­sel wall move­ments in order to cre­ate dig­i­tal twins of the heart. They must be biofi­del­ic, i.e., must repro­duce both phys­i­o­log­i­cal real­i­ty and cer­tain patholo­gies, in order to be able to test med­ical devices. We will also need to col­lect the most com­plete data pos­si­ble from clin­i­cians. The use of AI will enable us to speed up cal­cu­la­tions and obtain results in a time­frame that is com­pat­i­ble with clin­i­cal and indus­tri­al real­i­ty’, assures Anne-Vir­ginie Salsac.

The sec­ond project, also sup­port­ed by Sor­bonne Uni­ver­sité, involves the LMAC lab­o­ra­to­ry for UTC, a lab­o­ra­to­ry renowned for its method­olog­i­cal con­tri­bu­tions. ‘This CAP con­cerns ener­gy stor­age. The aim is to car­ry out mul­ti-scale mod­el­ling in the design of Li-ion bat­ter­ies in order to opti­mise their per­for­mance and max­imise their lifes­pan. The meth­ods used to design bat­ter­ies of dif­fer­ent sizes are not the same. The idea in this CAP is to use applied math­e­mat­ics and, in par­tic­u­lar, par­tial dif­fer­en­tial equa­tions, to build mod­els that favour reli­able pre­dic­tive approach­es that off­set the high cost of gen­er­at­ing exper­i­men­tal data’, explains Sal­im Bouzebda.

MSD