C2RO is mak­ing ro­bots smarter us­ing the cloud

Montreal Gazette - - EXTRA - jsere­brin@post­media.com

Ro­bot-mak­ers have a chal­lenge, said Ric­cardo Badalone.

They want to take ad­van­tage of re­cent devel­op­ments in ar­ti­fi­cial in­tel­li­gence and make their ro­bots smarter.

They could do that by re-en­gi­neer­ing their ro­bots and build­ing more mem­ory and more com­put­ing power into them. But that’s ex­pen­sive and mem­ory-in­ten­sive ap­pli­ca­tions re­quire more bat­tery power, which could make mo­bile ro­bots less use­ful, since they would have to charge more fre­quently.

Badalone is the CEO of C2RO, a com­pany that says it has a so­lu­tion: It wants to put ro­bot brains in the cloud.

The com­pany has de­vel­oped a sys­tem to process sen­sor data from ro­bots in the cloud in real time, said Soodeh Farokhi, the com­pany’s CTO. It then sends con­trol com­mands back to the ro­bot.

That means the ro­bot doesn’t need to have com­pli­cated com­puter sys­tems on board, she said.

There are other ad­van­tages that come with the flex­i­bil­ity of cloud com­put­ing, Badalone said.

If ro­bot op­er­a­tors want to add mem­ory to their ro­bots, they can do that re­motely, with­out the need to phys­i­cally up­grade their ro­bots. If they’re not us­ing as much mem­ory, even tem­po­rar­ily, they can re­duce it just as eas­ily, he said.

“You can in­crease the ca­pa­bil­i­ties of the ro­bot in­stantly and with 100 per cent flex­i­bil­ity,” he said.

That ex­tra pro­cess­ing power is im­por­tant for ro­bots that are used in­doors. (Ro­bots that do spe­cific job in­doors — like ro­bots that clean or trans­port things — as well as those that in­ter­act with peo­ple are the com­pany’s main fo­cus.)

Those ro­bots strug­gle to lo­cate them­selves, Farokhi said, be­cause there’s no GPS for in­side build­ings.

“The way that we do it, we use vis­ual SLAM, si­mul­ta­ne­ous lo­cal­iza­tion and map­ping,” she said.

It’s a tech­nique where ro­bots use on-board cam­eras to cre­ate “a 3D map of their en­vi­ron­ment in­doors, and they can au­tonomously nav­i­gate them­selves to un­der­stand the en­vi­ron­ment and go from A to B.”

But that re­quires sig­nif­i­cant com­pu­ta­tional power.

Con­nected ro­bots could also act based on data from sen­sors and cam­eras that are off-board as well as to col­lab­o­rate.

“When we have the map of one ro­bot in the cloud, the other ro­bots can use that map,” Farokhi said.

For ro­bots that work to­gether — to clean a large area, for ex­am­ple — that can be ad­van­ta­geous.

That’s not the only rea­son C2RO wants to put that ca­pa­bil­ity in the cloud, though.

Con­nected ro­bots could also be given new ca­pa­bil­i­ties through ar­ti­fi­cial in­tel­li­gence mod­ules — pieces of soft­ware that give the ro­bots spe­cific ar­ti­fi­cial in­tel­li­gence­based skills.

A mod­ule in the pro­gram that pro­cesses data gath­ered from a ro­bot’s sen­sors could gen­er­ate more us­able in­for­ma­tion from that data, for ex­am­ple.

“That’s why it’s so pow­er­ful to have it in the cloud, be­cause we can start com­bin­ing dif­fer­ent ar­ti­fi­cial in­tel­li­gence mod­ules to give the ro­bot more ca­pa­bil­ity,” Badalone said.

C2RO is cur­rently de­vel­op­ing sev­eral mod­ules in-house and, in the fu­ture, it plans to let other com­pa­nies and re­searchers de­velop their own mod­ules for ro­bots run­ning on its plat­form.

JOHN MA­HONEY

C2RO founder and chief tech­nol­ogy of­fi­cer Soodeh Farokhi, with CEO Ri­cardo Badalone. The com­pany seeks to take ad­van­tage of re­cent devel­op­ments in AI to make their ro­bots smarter.

Newspapers in English

Newspapers from Canada

© PressReader. All rights reserved.