As a uni­ver­si­ty fo­cused on us­ing re­search for the up­lift­ment of its com­mu­ni­ty, the Uni­ver­si­ty of the West In­dies (UWI), St Au­gus­tine, the pre­mier ter­tiary in­sti­tu­tion in the re­gion, pro­duc­ing world-class schol­ars, be­lieves sci­ence should be ac­ces­si­ble to the pub­lic. We of­fer this me­dia se­ries, UWI Sci­en­tists Speak, where our sci­en­tists, three of whom were re­cent­ly award­ed the high­est award of the na­tion–the Or­der of the Re­pub­lic of Trinidad and To­ba­go, will present some of their work.

This week, we hear from Dr Craig Ram­lal on his work and the work done at the Fac­ul­ty of En­gi­neer­ing us­ing ar­ti­fi­cial in­tel­li­gence (AI) to solve re­al-world prob­lems.

Dr Ram­lal leads the con­trol sys­tems group and co­or­di­nates the post­grad­u­ate stud­ies and re­search for the De­part­ment of Elec­tri­cal and Com­put­er En­gi­neer­ing at The UWI, St Au­gus­tine Cam­pus. He serves as the prin­ci­pal in­ves­ti­ga­tor of the In­tel­li­gent Sys­tems Lab and is a tech­ni­cal ad­vis­er to re­gion­al AI-re­lat­ed poli­cies.

—Prof Rose-Marie Belle An­toine, prin­ci­pal, UWISTA

As thought lead­ers, it is very im­por­tant that we give back to our com­mu­ni­ties and that our stu­dents take note of our ac­tiv­i­ties. I be­lieve it is more mean­ing­ful for stu­dents to learn from our ex­pe­ri­ence and ben­e­fit from the world we are try­ing to build, and when it is their turn, they al­so fol­low suit. Re­cent­ly, my re­search has been aligned with in­tel­li­gent con­trol strate­gies, game the­o­ry, and ar­ti­fi­cial in­tel­li­gence. I am en­gaged in this field through three key av­enues.

First­ly, through ac­tive in­volve­ment in craft­ing in­tel­li­gent con­trol, learn­ing, and rea­son­ing al­go­rithms, we seek to un­der­stand how these sys­tems per­ceive and in­ter­act with the world. Sec­ond­ly, by ap­ply­ing these al­go­rithms to ad­dress re­al-world chal­lenges in in­dus­tries and ed­u­ca­tion­al en­vi­ron­ments. Last­ly, through ded­i­ca­tion to shap­ing poli­cies and reg­u­la­tions that en­sure the re­spon­si­ble and se­cure im­ple­men­ta­tion of these tech­nolo­gies.

In 2019, Dr Arvind Singh and I co-led the ARC Labs team of UWI, part­ner­ing with re­searchers from Tallinn Uni­ver­si­ty and Es­to­nia’s na­tion­al pow­er com­pa­ny Eler­ing AS to de­vel­op deep-learn­ing ob­ject de­tec­tors for pre­dic­tive as­set man­age­ment of their elec­tri­cal grid. We suc­cess­ful­ly de­vel­oped al­go­rithms that could ac­cept ex­treme­ly large res­o­lu­tions of fly-by im­age da­ta and process it to give a Health In­dex of the elec­tri­cal in­fra­struc­ture. This en­ables the grid op­er­a­tors to de­ter­mine where to spend ef­fort and re­sources be­fore a black­out can oc­cur.

In 2020, I was part of UWI’s En­gi­neer­ing COVID-19 re­sponse team to de­vel­op so­lu­tions to mit­i­gate the spread of the virus. We were tasked with de­vel­op­ing face masks, face shields, pro­tect-a-doc­tor kits, and splash box­es. I was the prin­ci­pal in­ves­ti­ga­tor for de­vel­op­ing ven­ti­la­tors, CPAP de­vices, air de­con­t­a­m­i­na­tion sys­tems, and UV sani­ti­sa­tion ro­bot­ic sys­tems. The ro­bot­ic and de­con­t­a­m­i­na­tion sys­tems used ma­chine learn­ing al­go­rithms to de­ter­mine clean­ing reg­i­ments based on room and air­flow pa­ra­me­ters.

 We worked with re­searchers from the Uni­ver­si­ty of Flori­da and of­fi­cials from the Min­istry of Health, with fund­ing from the Na­tion­al Gas Com­pa­ny (NGC) and the Cana­di­an High Com­mis­sion, to de­vel­op and im­ple­ment these sys­tems. Through this process, we es­tab­lished stan­dards, re­dun­dant sup­ply chains, and in­dus­tri­al link­ages to rapid­ly de­ploy pro­tec­tive sys­tems against fu­ture pan­demics.

In 2022, I part­nered with re­searchers from Rut­gers Uni­ver­si­ty to de­vel­op col­lab­o­ra­tive om­ni­di­rec­tion­al ro­bots ca­pa­ble of trans­port­ing ob­jects on a de­formable sheet via op­ti­mi­sa­tion and re­in­force­ment learn­ing al­go­rithms. The ap­pli­ca­tions of this tech­nol­o­gy are wide­spread as the ro­bot­ic team can lift any shaped ob­jects, ro­tate, and trans­port them. This re­search paved the way for ex­plor­ing ar­eas like at­tack re­silien­cy, giv­en the vul­ner­a­bil­i­ty of col­lab­o­ra­tive au­tonomous sys­tems to wire­less hack­ing. From this, my post­grad­u­ate re­search stu­dents are in­ves­ti­gat­ing pla­toons of elec­tric au­tonomous ve­hi­cles and the in­tel­li­gent con­trol al­go­rithms that en­hance their re­silience to cy­ber­at­tacks.

Large lan­guage mod­els (LLMs) are now well-ac­cept­ed by the pub­lic and in­dus­try. I have been in­volved in ma­chine rea­son­ing and sep­a­rate­ly large lan­guage mod­el re­search for some time, de­vel­op­ing our LLMs and mak­ing them small­er and more in­tel­li­gent. Part of my work is to bet­ter ed­u­cate our na­tion’s work­force to use these tools re­spon­si­bly and safe­ly and un­der­stand their lim­i­ta­tions. This is nec­es­sary to boost pro­duc­tiv­i­ty and to keep glob­al­ly com­pet­i­tive. In that trend of thought, with fac­ul­ty col­leagues, we have de­vel­oped short cours­es through our de­part­ment tar­get­ed at a wide range of ed­u­ca­tion pro­files. The in­dus­try has been keen on the adop­tion of this tech­nol­o­gy. I have worked with or­gan­i­sa­tions to im­ple­ment these sys­tems in their su­per­vi­so­ry con­trol and da­ta ac­qui­si­tion, alarm, and con­trol sys­tems.

These LLMs al­so im­pact ed­u­ca­tion re­search and its fu­ture. My work in this realm has been two-fold; in the im­me­di­ate case, it re­volves around how ed­u­ca­tion in­sti­tu­tions should re­act to stu­dents’ use of LLMs to en­sure in­tegri­ty and hon­esty in stu­dents’ sub­mis­sions. Sec­ond­ly, I ex­plore what the fu­ture of teach­ing could look like with rea­son­ing ma­chines. In this case, the LLM al­go­rithms and in­te­grat­ed sys­tems are mod­i­fied to sup­port ped­a­gogy and the course’s learn­ing out­comes rather than the cur­rent glob­al norm, which is the use of LLMs as a tool with­in the course.

The lat­ter, if not done well, can sig­nif­i­cant­ly de­grade the qual­i­ty of teach­ing. To this end, with an in­ter­de­part­men­tal group at UWI, we have de­vel­oped a nov­el AI-en­abled soft­ware plat­form for the ap­pli­ca­tion of dig­i­tal twins. In this work, AI aids in train­ing, de­ci­sion sup­port and de­sign­ing con­trollers for ro­bot­ic, pow­er, and in­dus­tri­al sys­tems. This tool can help both in­dus­try and stu­dents to study sys­tems and their be­hav­iours.